Lets assume that there is a language that has some relatively small set of meaning blocks, within 10-1000, so that any word can be composed from this meaning blocks. Each meaning block should have only one own meaning, not a set of them. By combining 2-5 of such meaning blocks we could achieve the same amount of options as exist in modern languages.

Idea is to define this set of meaning blocks once and do not add them later, but use their combinations if a new word is required.

Is there a name for such experimental language, or any similar works done about this topic?

(for now im not interested in grammar, words order.

rules for composing words from this meaning blocks and the list of the blocks themselves is what i want)

upd1: NSM is a very close to what i was looking for. each word has only one wide meaning, but not a set of them. But probably there are too few meanings in NSM and i dont understand their logic in selecting 'above' but not 'in front' or 'left' for example. Or lack of color, noise, tactile feelings details. Also many prople are against this theory, whivh i didnt meet with toki pona, so probably there is something wrong about NSM that i do not see yet.

Toki pona has a big community, but it meanings associated with words are unconnected. Each word has many separate meanings, and this is the biggest problem for me. Also some important meanings are not present and many of unimportant ones are present. like the meaning 'part' is abscent but 'insect' and 'non-cute-animal' are present. So again, i dont understand the logic in selecting this particular words.

List of requirements(its personal, but im trying to be objective): 1) each word has connected, close set of meanings. So that each meaning is as close as synonym to any other meaning of the same word. NSM does fit, toki pona doesnt. 2) all important meanings are present, or logic in selecting existing meanings is explained (i didnt it find yet not for NSM, nor toki pona) 3) unimportant meanings are not present (toki pona does not fit, meanings like 'insect', 'non-cute-animal' and 'powder' can be composed from other meanings, but this is personal) NSM does fit 4) active community. Toki pona does fit well. But with NSM people do know about it but are actively agast it(i dont know why), so NSM does not fit.

So probably there is some other option apart from this two?

  • 2
    What is a meaning block?
    – user6726
    Jan 25, 2017 at 20:49
  • 2
    What is your metric of simplicity? For example, is "dog" a meaning block? Is "square" a meaning block?
    – user6726
    Jan 25, 2017 at 21:09
  • 2
    So "simplicity" is irrelevant, what matters is "irreducibility". Are "life form" and "shape" meaning blocks? Why isn't "life form" reduced to more general "entity", or is it?
    – user6726
    Jan 25, 2017 at 21:52
  • 2
    "Each meaning block should have only one own meaning" — sounds almost impossible. Try defining "hammer", and you will end up with zillion different words for zillion various hammers made of different materials (and their combinatorial compositions), shapes (and compositions), colors (and compositions), and so on. Immediately at the moment you decide to stop, you get an ambiguity of a single (already huge) word having two distinct meanings that differ by a property you haven't listed yet. Jan 25, 2017 at 21:53
  • 3
    The term is "semantic primitive" (or "semantic prime") and there is a body of lit on it. search "natural semantic metalanguage". unfortunately is does not work.
    – mobileink
    Jan 26, 2017 at 19:45

7 Answers 7


I'll just drop a list of some philosophical languages dedicated to a strictly compositional system making use of such "primitive meaning blocks", there have been many such attempts (there was a hype among philosophers around the 17th century):

1) Args signorum is a language constructed by George Dalgano in 1661.
Quoting from the FrathWiki entry:

Words in Ars signorum are formed by derivation and compounding from a set of 1068 monosyllabic roots ("Radicals"). These roots denote generic concepts and are derived from a taxonomy of ideas. The classes indicated by the initial phoneme are:

  • A being.
  • Η substance.
  • E accident.
  • I complete (concrete) being.
  • O body.
  • Υ spirit.
  • ...

Example of root formation:

  • neik terrestrial animal
  • nak oviparous (e.g., lizard)
  • nηk whole-footed (e.g., horse, elephant)
  • ...

This should come close to what you wanted.

2) aUI is a (relatiely new, created in 1952 by the philologist and psychoanalyst Dr. John Weilgart) conlang intended to serve as a universal (not only world-wide, but universe-wide) lingua franca using a similar approach of

semantic primes (the simplest building blocks of meaning, ideas that cannot be broken down any further). [...]
In creating combinations, we take the necessary and sufficient characteristics in a nutshell.

You can check out their online presence, they have a nice introudction + bidirectional dictionary.

3) In An Essay towards a Real Character and a Philosophical Language, John Wilkins (1668) proposes a similar tree-like ontology (and actually uses symbols directly depicting relations in this tree structure system, which were later also assigned a phonetic system):

Wilkin's "Real Character" is a constructed family of symbols, corresponding to a classification scheme developed by Wilkins and his colleagues. It was intended as a pasigraphy, in other words, to provide elementary building blocks from which could be constructed the universe's every possible thing and notion. [...]
Later in the Essay Wilkins introduces his "Philosophical Language," which assigns phonetic values to the Real Characters.

4) In 1678, Gottfried Leibniz proposed lingua generalis.
The basic idea was to assign every atomic property a prime number, and complex meanings are then computed as the arithmetic prodcut of these numbers. (Since prime numbers are chosen, any such product is uniquely identified by its atomic properties. However, this means that applying "reverse engineering" to identify the atomic factors boils down to integer factorization which is a rather complex mathematical operation and, as a consequence, not that well suited for practical applications...).
I'm not sure, though, whether Leibniz actually computed such a list of atomic properties; I couldn't find any good reference online.

5) Blissymbolics is

a semantic graphical language that is currently composed of more than 5000 authorized symbols [...]. It is a generative language that allows its users to create new Bliss-words as needed.

Although, as the other examples, originially inteded as a univeral language aiding culture independent communication, it is nowadays

used by individuals with severe speech and physical impairments around the world.

6) Last but not least, Solresol is a musical language employing a similar, though less strictly compositional way of composing meaning1:

In Solresol morphology, the longer words are divided into categories of meaning, based on their first syllable, or note. Words beginning with 'sol' have meanings related to arts and sciences, or, if they begin with 'solsol', sickness and medicine (e.g. solresol, "language"; solsolredo, "migraine").

Unlike for all of the other artificial languages, I know someone in my circle of acquaintances who is actually "speaking" this language to a certain extend; there seems to be a mildly active community that actually uses this conlang.

1 What I find the most interesting feature about this language is that

meanings are negated by reversing the syllables in words. For instance fala means good or tasty, and lafa means bad.

This is what we used to do as kids when we decided it was opposite day. To a question "Do you like ice cream?", you were supposed to reply with "No, of course not, there is nothing I hate more than sweets", however it was perfectly valid to say "seY" instead.
With all the weird stuff that happens in natural languages' phonology, I have never found a language that uses phonetic reversal this systematically - probably because it completely destroys the nice compoistionality.

  • Thats a lot of languages! thank you. Did spend some time trying to find main symbols of blissymbolics, here it is, if somebidy needs: the-symbols.net/blissymbolics/dictionary - probably the most interesting one, but i cant find just a key symbols separately. Languages 1,2,3,6 are using about 20 basic concepts to try to fit 1 concept = 1 letter, thats probably too few. but their combination just from 2 symbols may be just right, need to check. Prime numbers language seems fundamentally wrong as 2+3=5 and 2+3+5+7=17, but 2^n numbers will work as 1+2+4+8=15(cant reach 16)
    – K Oliver
    Jan 28, 2017 at 10:23
  • @K Oliver Concerning the prime numbers: 1) Leibniz uses multiplication, not addition, I think it would not work otherwise, 2) That you can't reach all numbers is not "fundamentally wrong", but the whole point of it, because 16 would not have a unique meaning, but two possible combinations (16 = 2 x 8 = 4 x 4) which was intended to be ruled out by using prime numbers as primitive factors. Jan 28, 2017 at 10:38
  • assuming no doubling 16 has only 1 meaning =16. and 15 hss only 1 meaning 1+2+4+8. And in binary numbers you can see each meaning (bit set to 0 or 1) without calculations. Multiplication with prime numbers will work, but not any better than binary numbers. assuming you want to store 10 first meanings. then prime product will be 2*3*5*7*11*13*17*19*23*29=6469693230, while the same for binary numbers is 1+2+4+8+16+32+64+128+256+512=1023. both are unique, but second case needs much less memory and computations. The only advantage of prime numbers prod is that you can store a number several times.
    – K Oliver
    Jan 28, 2017 at 11:27

toki pona is a constructed language which is very close to what you're looking for, it has only 120 words and you cannot add any new ones, you can just combine the existing ones to get the meaning you need. Still the minimalism of toki pona is the point of that language, you're supposed not to construct complex structures of the 120 words it has, but rather use them in their massively wide meanings, yes, unfortunately words in this language have very wide meaning. Anyhow, you can judge it yourself if it's what you mean, here's the official toki pona textbook.

  • 2
    Huh. So someone actually decided to implement this nutty idea. I tried to render the sentence "I ordered the goat stew, not the lamb curry", and failed. Guess the Toki Pona people live very simple lives.
    – user6726
    Jan 26, 2017 at 1:24
  • 3
    @user6726 - It's all a matter of convention. You can decide it and agree it with you friends that you call a goat, say, 'soweli telo mama lili' (little parent liquid animal >> little milk animal), and a lamb 'soweli linja len lili' (little clothes hair animal >> little wool animal). 'Stew' can be 'moku telo soweli' (animal liquid food >> meat liquid food) and 'curry' can be 'moku kasi wawa' (intense plant food >> spicy food). And naturally, 'I ordered' is 'mi mani'. ))
    – Yellow Sky
    Jan 26, 2017 at 2:34
  • 1
    In toki pona each word has a qiite big meaning on its own and the words arent easy to compose together to make a more complex meanings. And some important words lile part/whole and left\right are missing while some very specific words like non-cute-animal or powder are present.
    – K Oliver
    Jan 26, 2017 at 6:32
  • So in toki pona each word can have a number of meanings, i.e. a number of different collections of semantic atoms?
    – Mitch
    Jan 26, 2017 at 22:21
  • 1
    Why, yes, I have. I would have gone with "carries footballs animal", but YMMV.
    – user6726
    Jan 28, 2017 at 0:55

this may work as an example of such meaning blocks



There are two related theories of lexical semantics that seem amenable to this kind of thinking.

One is feature semantics, where a word can be assigned a number of positive or negative features (by analogy with phonemes having phonological features). A word's semantics is defined by a selection of binary features so 'man' might be [+human][-female] and 'table' might be [+furniture][+flat top][+has legs]. If an item doesn't have a positive or negative version of a feature, then that feature is just not relevant.

Another model is 'ontology'. This is not the philosophical ontology (metaphysics of being) but rather the specification of semantics by relation to other entities. You can think of any concept as a node and how it is related to other concepts by labeled arrows to other nodes.

(you can implement feature semantics in an ontology by having all male things point to a 'male' concept)

You ask if there is a finite set of things here from which the seemingly infinite set of all words or concepts can be derived. Mathematically, yes, this could be done. With a set of 100 features, there are 2^100 (way more than the number of physical atoms in the universe). However I don't think this is possible usefully. Real life has so many possibilities to it, new things appear all the time. There might be a finite number of arrow labels ('is-a', 'has-a', 'is similar to', etc), but it seems reasonable to think that there are no terminal nodes (no outgoing arrows); there could easily be cycles.

Just to mention quite an opposite direction, axiomatics in mathematical logic depends on treating axioms as the atoms for which there is no proof. All proofs bottom out at these axioms. Whether you consider them syntactic or semantic atoms is another discussion.


Such kind of language would be a constructed language.

There are two different types of constructed languages.

  1. Naturalistic languages (like Esperanto or Interlingua) that follow closely the features of natural languages and that are usually intended as means for international communication or serve a role in pieces of art (as the Elbian languages of Tolkien or as Klingon in Star Trek).
  2. Philosophical languages or a priori languages (Famous example: Solresol, also Leibniz tried to construct one) that are trying to create a system of communication from some kind of first principles.

Your experimental language would fall into the second category.


Surprisingly, some human languages are very close to this: https://kanjialive.com/214-traditional-kanji-radicals/ kanji, the japanese written language, sorted from fewer lines(simple) to a more lines (more compkex symbols)

Aparently all symbol wriying systems, from egypt to japan do more-or-less fit into symbol=meaning category of languages.

Altrough its a bit archaic and simple symbols represent what was popular long time ago, like animal types and weapons, and meanings may be disconnected, its still an interesting option to 'beef up' some artificial language by taking something from this writing systems.


As mentioned by others, this would likely be a form of philosophical constructed language.

But this also bares similarities to polysynthetic languages, where words are constructed from stringing together various morphemes. Or to the theoretical idea of an Ogliosynethic Language, this however does not exist naturalistically and only exists in constructed languages.

It is also similar, at least in writing, to Chinese Hanji/Japanese Kanji. Where there are various symbols that could be strung together. These, however, have multiple meanings per symbol, therefore would not fit the requirements of what you are asking for.

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.