I started wondering what share of all world's languages are polysynthetic (on any practical definition of polysynthetic, i.e. the prototype approach, the macroparameter theory (Baker 1995), etc.), and I can't find even a rough estimate. Does anyone know?

Some other parametric values do have counts, for instance, you often learn that SOV languages constitute almost a half of world's languages. But nothing like that for isolation, agglutination and polysynthesis?

  • You may want to take a look at linguistics.stackexchange.com/a/1078/445
    – Alex B.
    Commented Nov 10, 2015 at 0:09
  • and another post linguistics.stackexchange.com/a/1088/445
    – Alex B.
    Commented Nov 10, 2015 at 1:25
  • 1
    @AlexB.: Thanks for the links! But you're not saying polysynthesis is now a mere unhelpful idealistic illusion, are you? Commented Nov 10, 2015 at 5:58
  • 1
    I'm not quite happy with the agnostic trend that I see in the discussion so far, so I edited trying to specify that I mean: any can be adopted among the various definitions of polysynthesis. There are quite a few, because (some) linguists recognize polysynthesis as a thing. Commented Nov 10, 2015 at 6:13
  • 1
    @AlexB.: That's right -- it's one of the most common (empirical) critiques of his theory that the definition excludes some languages that are generally considered polysynthetic. That said, even that approach would do for my inquiry. Commented Nov 11, 2015 at 13:15

3 Answers 3


I think the problem lies in the "on any practical definition" part. After decades in the field, I still don't know what the difference between agglutinative and polysynthetic is. Imbabura Quechua is an "agglutinative" language with a small morphology, and Sanskrit is an analytic language with a big morphology. Counting morphemes doesn't help, you have to have more sophisticated metrics of "combinability" and "functional uniqueness". Even if such notions were computationally defined, processing data from languages would require a lot of work, compared to what's required to determine if the dominant word order is SOV.

  • Well yes, and still most, if not all, linguists will agree that Turkish is agglutinative, and not polysynthetic, but Mohawk is polysynthetic (whatever agglutination there might be). I imagine someone could apply their own understanding of morphosyntactic type and count the languages on that understanding. Maybe I should :-) And isolating languages are less controversial, and nobody count even them! Thanks for your answer though. Did/do you work on Imbabura Quichia? Commented Nov 9, 2015 at 22:50
  • I did a tiny bit of work on IQ, and was adjacent to such work. Before we count languages, I would want to know that we're invoking a meaningful distinction (not just a definable distinction). For example, we could define a "type" of language as "gloofy" if the language has a causative affix and a sg/pl object distinction on verbs, and non-gloofy otherwise. But that would be pointless. I don't (generally) see what the utility is of making the distinction.
    – user6726
    Commented Nov 9, 2015 at 23:48
  • 1
    @IvanKapitonov Without intending disrespect to anyone, isn't it somewhat irrelevant what "most linguists" think about this, as long as specialists in (putatively) polysynthetic/agglutinative/etc. languages are unsettled on the question? Most professional linguists probably know that Mohawk and some other languages are traditionally termed polysynthetic, but that is not the same as being conversant (even passingly) in any of these languages.
    – user8017
    Commented Nov 10, 2015 at 2:44
  • 1
    Many of us linguists have polysynthetic languages in our checkered pasts. I have Lushootseed, for instance. They're sought out because they are such extreme cases. But they're stable; all the languages in the Northwest Coast Sprachbund are polysynthetic, even though there are half a dozen language families involved. They all have similar -- and similarly extreme -- phonologies, as well.
    – jlawler
    Commented Nov 10, 2015 at 3:26
  • 1
    @jlawler, I probably should have proposed Lushootseed as a case in point. I would not have called Lushootseed a polysynthetic language -- so why do you? I'm willing to expand the polysynthesis empire, but I wanna know what properties you're referring to.
    – user6726
    Commented Nov 10, 2015 at 5:34

I won't answer this directly but rather give resources that I think can answer it.

The WALS (World Atlas of Language Structures) has three chapters and accompanying maps on inflection types, 20, 21, and 22. They have a broad inventory of languages.

They do not use the same terminology as the classes agglutinative, isolating, etc. but do start from analytic and synthetic and refine it from there.

But from those maps you can tell frequency almost directly; they give the number of languages in their inventory with each particular feature, with all the limitations of such an inventory (selection bias, essentialism (lack of vagueness), your favorite may not be in there, etc).

You'll notice there there is no single map for SVO permutations either.


Originally polysynthesis only meant that an average word had a high content of morphems or meaning elements, without any clear definition how high it should be to call a language polysynthetic.

Later on one tends to call languages that are head marking for at least subject and object as polysynthetic, and one will often add the criterium of noun incorporation. But even if a language has these traits, it does not necessarily have a high content of elements in an average word.

I would define a polysynthetic language as one with both these traits.

Remark that a headmarking language can be dependentmarking at the same time, also called doublemarking, and of cource still be polysynthetic.

Agglutination or fusing of elements have nothing to do wuth the definition, exept that a polysynthetic language must at least be halfly agglutinating to be learnable.

It is very difficult to answar how common this language type is, because many languages have structures that are not so easily classified.

Italian, French and Spanish, for example have a higly templatic verbal group consisting of the main verb, auxiliary verbs, object clitics and adverbial clitics, and these languages are partially headmarking. And these clitics behave more like affixes than just clitics.

So these languages are at least very near to be polysynthetic, and the same holds for modern Greek.

There is also a bias against calling traditional culture languages, like Hungarian, Georgian, Romance languages and Modern Greek polysynthetid, because that would be the same as classifying them togeather with the languages of peoples tagged as primitive.

  • The WALS database has no polysynthetic or agglutinative categories, preferring rather to unpack these ideas into several precisely-defined features. When I tried to use a set of features to approximate "agglutinative", my selection ended up including English... un-frig-ing-believ-able! Commented Sep 30, 2016 at 15:53

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.