Would there be a theoretical way to distinguish between two related languages which diverged a long time ago, like tens of thousands years, and two totally unrelated languages, say, created independently?

I must say that I realize this question is still probably a big topic in research but I'm curious if there's any possible answers or else.

To be more specific, as lemontree suggested, my question should be interpreted as "Up to how many thousands of years is reconstruction still possible?" (cited from lemontree's comment) or "Are there any features that are very persistent throughout language change and could give sound clues to relatedness?".

  • "A theoretical way to distinguish them"? Well, you can simply say that there are related and unrelated languages and you got your distinction. Or do you mean whether it is in general possible to tell a remotely related language from a totally unrelated one, even if the related language diverged a long time ago, i.e. to which degree/after how much time commonalities are still visible even after early having split up into different directions? Dec 1, 2016 at 18:49
  • Yes this is what i meant. Wasn't my question clear enough?
    – user14639
    Dec 1, 2016 at 19:15
  • 1
    The wording "theoretical" confused me, because it sounded like you wanted to know if there was anything apart from straightforwardly comparing the languages and looking for similarities in lexicon, syntax etc. Dec 1, 2016 at 19:28
  • 3
    The maximum time depth for the comparative method -- the best one we have -- is between 6000 and 8000 years. That is, we're pretty sure about a lot of language families that have left records -- Mayan, indo-European, Chinese -- because the records extend the depth. We have I-E records from about 3500 years ago, so that allows the comparative method to look back to about 5-6000 BCE and say that Indo-European languages have a common ancestor. But never tens of thousands of years; that's impossible. The question of relation is meaningless at that time depth.
    – jlawler
    Dec 1, 2016 at 21:02
  • 1
    Thanks for the beginning of answer. But why is it impossible? And why would it be meaningless?
    – user14639
    Dec 1, 2016 at 21:39

1 Answer 1


There are methods for calculating the age of the split between two languages (which, conveniently, I do not rembember the details of). They rely on counting related material versus unrelated material, and making a proportion. Evidently, there is a point at which such proportion tends to zero. That is, two closely related languages that diverged a thousand years ago will have x% related words; two not so closely related languages that diverged two thousand years ago will have x/2% related words, and so on. But at some point there will be no difference between relational similarities and statistical noise.

Once in a conlang forum two groups invented different vocabularies for two unrelated languages. I applied the usual method and concluded that the languages would have diverged 5,600 years ago. So it is possible to conclude that two completely unrelated languages are close enough that they have diverged five or six millenia ago; that probably is close to the point where the difference between signal and noise becomes untraceable.

  • 1
    Glottochronolgy. But it is controversial, and even if it is accepted in theory, I don't believe there's any consensus about the numbers.
    – Colin Fine
    Dec 3, 2016 at 23:45
  • I understand that the level of noise increases with time. This kind of answers my question although, now I am wondering if there is a precise model for the amount of noise that occurs with time. And if we are able to reconstruct past language based on modern ones, how much information has been lost from the reconstructed language?
    – user14639
    Dec 4, 2016 at 10:08
  • @TTFarreo The most known models are those of Swadesh and Starostin. As Colin Fine points out, Glottochronology (see en.wikipedia.org/wiki/Glottochronology for very basic information) is controversial. And, as far as I know, "reconstructed language" is an overstatement. We wouldn't be able to speak Proto-IE, or understand it if we had a TARDIS and went to Bronze Age Eastern Europe. So, lots of information were certainly lost. Dec 4, 2016 at 11:17

Your Answer

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