I've read several studies about sound symbolism and I'm still not sure whether I got an insight into the topic. I know that today's view of most of the linguists is skeptical towards sound symbolism to the extent that most of the words in the most of the world languages do not resemble the sound of the word but there are many antonyms that could resemble their meaning in the contrast of the neurological or the vocal human cognition system. Though, even though we could find resemblances between the words and their meaning in our studies, their interconnectedness still might be theoretically a result of the shared "convention", an old trait of language perception shared among the earliest languages.

But let's say a study found a significant above-random correlation between words' meanings and the meanings guessed by people from two very distant language groups. Is there a non-negligible chance that this correct guess could be actually caused by the shared root of the proto-human language?

  • Do you mean "some two people" or "any two people"? The latter (a.k.a. "cherry-picking") would be significant, the former would not.
    – user6726
    Dec 15, 2016 at 16:21
  • @user6726 Neither of those options. I mean "any two groups of people", but yes, I the purpose of the theoretical experiment would be to find the roots (therefore any) of our languages not to find it's possible to find similarities.
    – Probably
    Dec 15, 2016 at 16:25
  • In other words, the experimental plan would be to randomly pair speakers of unrelated languages and see how often they can discern that "koira" and "mbwa" mean "dog". I'm trying to understand the "if" part ("found a significant correlation").
    – user6726
    Dec 15, 2016 at 16:33
  • 1
    You're confusing sound symbolism with onomatopoeia. That's a common mistake. Onomatopoiea occurs only with words that refer to noises. Not many words do, so onomatopoeia is rare. Sound symbolism, on the other hand, is very very common in most languages, though sound-symbolic meaning doesn't work the same way as ordinary morphemes do. For instance, check out the English simplex words beginning with /kl-/.
    – jlawler
    Dec 15, 2016 at 21:01
  • 1
    Neither onomatopoeia (words for sounds that sound like what they mean, like bell) nor sound symbolism (systematic and widespread meaningful attribution to individual speech sounds or clusters) is a "theory of the origin of language". They are simply phenomena of language that go back as far as we have records of language and probably should be presumed to be available earlier. But there is no "the origin of human language", any more than there is a "the origin of the human thumb". Both owe features to phenomena of other species that gradually changed to what we now have.
    – jlawler
    Dec 15, 2016 at 23:50

2 Answers 2


This part seems to have some presuppositions that need to be corrected:

Is there a non-negligible chance that this correct guess could be actually caused by the shared root of the proto-human language

First, it is possible that there was a single earlier language from which all current languages derive, but it is also possible that there is no single common ancestor to all human languages. Let's just assume that there was one proto-language, then what would it mean for a shared root to "cause" significant recognizability in words? We can take cases where having a "shared root" does affect guesses. Speakers of Germanic languages can guess the meaning of the word "hand" in another Germanic language because the words of Dutch, English, German, Danish, Swedish, Norwegian and Icelandic are derived from the same root of proto-Germanic. Likewise, speakers of Bantu languages can guess the meaning of the verb "to cultivate" in another Bantu language, because they nearly all have the same word pronounced [kʊrɪma] or something close to that. In such cases, you can say that ability to guess meaning correctly is caused by the shared root of Proto-Germanic or Proto-Bantu. But that implies something very weird: that speakers of English or Norwegian have some kind of mystical access to a fact of a language spoken millenia ago, which nobody alive now has ever heard.

What allows speakers of Germanic or Bantu to make these "educated guesses" is not the fact that the modern words historically come from a common root, it is the fact that the modern words look very similar (which itself is because they derive from a common foot, and because the sound changes that affected these particular words happened to have not changed the shape of these roots all that much).

When the modern words are not similar in shape, people are unlikely to correctly guess the meaning of a word in another language, thus speakers of Indo-European (who aren't bilingual and haven't studied Indo-European linguistics) will not guess that koun, ci, hound, chien, šuõ, šun, sobaka all mean the same thing. The root *gʷʰen is common enough in Indo-European languages, but people generally can't guess that because the contemporary manifestations are so diverse (defense, gìnti, gon, gjanj, φόνος, ganem). Even if a pair of words does historically derive from a common source, that fact doesn't cause people to correctly guess the meaning of the words.

There is a very small set of words whose meanings are cross-linguistically guessable at a somewhat significant rate, especially "mother", "father", "dog", "chicken". The problem is that these words also have a tendency to be irregular from the perspective of historical sound changes – they tend to resist regular changes, as though there is some sound-symbolic force that keeps them exceptionally close to the standard of mama, dada/baba, bu, kuk, to the point that it would be impossible for such similarity to be preserved even if Proto-Human had these roots which were passed down to most modern languages. It's that sound-symbolic force, and not a common root in a language possibly spoken 100,000 years ago, that causes good-guessing behavior for these words.

  • 1
    The cross-linguistic similarity of words such as those for ''mother'' and ''father'' is discussed here. Dec 16, 2016 at 2:59
  • Thank you, these words are prettymuch what I was looking for. Though the beginning talks about something else because those kinds of cases (bilingual people, languages from a similar language group) are exactly what I've refused I the comments.
    – Probably
    Dec 16, 2016 at 7:09
  • The point that you should give thought to is that English speakers cannot guess word meanings of German, in cases where the English and German words are highly unlike, even though the languages are closely related. The reason is, a close relation itself does not cause good guesses: what does, is surface similarity in form.
    – user6726
    Dec 17, 2016 at 17:04

I offer another perspective although I like the first answer.

The proto-Human language is not reconstructable. Two deep (and highly putative) language families that have been suggested by historical linguists are Niger-Saharan and Nostratic. Were these to be genuine, they would date to the early Holocene and Epipaleolithic respectively. Owing to great internal diversity, the Khoisan and Australian Aboriginal languages remain as strictly geographical superfamilies with no demonstrated genetic relationship between internal families.

There is a consensus in lexiostatistics that there is a meantime to happen for semantic shift. After a predicable number of years, a semantic value will be replaced with a new word. The replaced word isn't always recoverable. The meantime to happen is contingent on phonology and the semantic value of the word. Even the more stable items on an ordered Swadesh list (we, two, I, eye, mouth, louse) can't survive indefinitely. After about 9,000 years, most of these conservative words will have been replaced. Eric W. Holman, Aharon Dolgopolsky and Sergei Starostin were the chief linguists of this discipline, glottochronology.

To return to my examples, in Niger-Saharan (now completely abandoned) all that is reconstructable outside some typological features are the single phonemes of alleged noun classifiers. Nostratic is relatively better substantiated but not even that comes close to qualifying as as a language related to the proto-Human language.

As well as onomatopoeic words for mother, father and birds there are areal terms which travelled with prehistorical human technology. "Pan-African" glosses include */bVr/ for ashes and */um/ for nose. I former is most likely areal and the second onomatopoeic.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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