Tl;dr: What reasons do we have--besides glottochronology--to think that Proto-Afro-Asiatic is actually 14,000 years old?

So, if you know much about proto-languages, you might know that Proto-Afro-Asiatic (PAA) blows other proto-languages out of the water when it comes to antiquity. The most common figures I see cited in various works espouse a date of 14 or 15,000 years ago. No other proto-language of an uncontroversial language family even comes close to this; I've struggled to find any examples of securely dated proto-languages even half as old. Quite frankly, 14,000 years beggars belief for standard comparative methods, and from what I can gather, there's nothing structurally 'special' about PAA which would make it any more or less innovative than any other language family (the Proto-Semitic triliteral root-and-pattern system, which does structurally foster a particular kind of linguistic conservatism, was not original to PAA, and seems to have come about quite late.)

So I did some digging, and all I could find to back up that "14,000 year" figure was glottochronological estimates. I happen to be a skeptic of glottochronology in general; among other things, it tends to overestimate time depths in cases where vocabulary replacement takes place more rapidly and affects more supposedly 'stable' lexical elements due to periods of intensive interlanguage contact and/or multilingualism (Young Dyirbal being the most infamous example). And there is no doubt that several branches of Afro-Asiatic have substantial foreign components in their phonologies. (For example, recent research has been able to make the case pretty convincingly that PAA was not tonal, and that tone developed independently in Omotic, Cushitic, and Chadic languages as the result of extensive contact with tonal Niger-Congo and Nilo-Saharan languages.) So it seems plausible to me--likely--that the glottochronological figure could be severely exaggerated by mistaking vocabulary replacement due to interlanguage contact with vocabulary replacement due to slow language-internal change. Due to the generally restricted and specialist nature of Afroasiatic studies (and African linguistic studies in general), I do not know of any studies which have investigated loanwords in Afroasiatic languages from neighboring non-AA African languages.

I'm certain I can't be the only person who's ever gone, "14,000 years? Now that can't be right!". So do we have any non-glottochronological reasons to think that PAA is actually as ancient as is so often claimed? And if not, why has this figure has been touted so much in the literature without any qualification or caveat?

  • 1
    It certainly helps that we have really ancient extensive recordings of two languages of different branches of the Afroasiatic family, namely Akkadian (Semitic) and Ancient Egyptian (own branch). The time depth is less impressive when counted from the oldest attested members. Commented Jun 13, 2021 at 9:46
  • 3
    @jk-ReinstateMonica. Egyptian and Akkadian go back perhaps 5000 years maximum. Trebling that is still "impressive".
    – fdb
    Commented Jun 13, 2021 at 9:57
  • 2
    @jk-ReinstateMonica Egyptian and Akkadian are very old in attestation, absolutely. But PIE has access to very ancient data just about a millennia later than legible Egyptian from Hittite; Proto-Sinitic has access to data from Old Chinese; and so forth. And yet these reconstructions cannot project nearly as much depth in time from the date of first attestation as PAA dating estimates claim to do. PAA of course HAS to be very old—but 14,000 years remains a tremendously large number even when Egyptian becomes legible “only” in the 3rd mill. BCE.
    – Khove
    Commented Jun 13, 2021 at 16:16
  • 1
    @jk-ReinstateMonica The difference in age of attestation between Afroasiatic and Indo-European is only about one millennium. Egyptian becomes legible in 2690 BCE, Akkadian just a few centuries later, ~2400 BCE. Hittite shows up in ~1700 BCE, Vedic Sanskrit begins around the middle of that same millennium. We have two IE branches to compare against each other of comparable age to AA, yet PIE is not reconstructible even close to as far back. Even accounting for how old Egyptian and Akkadian must be, 14,000 years remains unheard of for anyother proto-language of an accepted language family.
    – Khove
    Commented Jun 14, 2021 at 1:52
  • 1
    Compare also this question and its comments and answers: linguistics.stackexchange.com/questions/13192/… Commented Jun 15, 2021 at 10:31

2 Answers 2


The basic reasoning is set forth in Ehret 2002, The civilizations of Africa : a history to 1800 in ch. 2. It is based on the claim that there was a civilization living in NE Africa, that they were the speakers of Afro-Asiatic (not Niger-Congo, Nilo-Saharan or Khoisan), and that their activities can be detected to some extent via archaeology. The physical evidence is mainly grindstones and blades, which archeologists have some reason to believe were used for certain purposes. In this book, he identifies a shift in grass-gathering practices in a part of the population of "Cataract peoples" between 11K and 13K BC, which he conjectures reflects the proto-Afro-Asiatic population. This new technology resulted in a significant expansion for that population. The reasoning becomes more linguistic when you look at what kind of words can be reconstructed to the proto-language: Ehret posits that there are common words for grains that might have been gathered, but not for livestock or cultivation, which then reasonably indicates that the proto-population didn't farm. Similar reasoning is used to at least start to hypothesize a date for proto-Bantu, given the reconstructability of farming and metal-working terms (therefore they did these things). In other words, the argument is not necessarily based on a computation of shared vocabulary or numbers of sound changes, it may be based on hypotheses regarding history based on archaeology and lexical reconstructability.

Unfortunately, there is no single scholarly work that purports to set out all of the archaeological and linguistic evidence that would support a specific date. AFAIK, Ehret has not addressed the alternative that the proto-culture spoke "pre-proto-Afro-Asiatic". The argument would crucially hinge on the lack for terms in the proto-language for activities known to have existed in that region at a certain date (farming). The discussion of bones and stones hasn't really addressed the question of "not later that X".

  • Brilliant, thank you so much! Ehret’s one of the more diligent I’ve seen about linking linguistic and archaeological evidence. I’ll have to look through this myself but it’s good to know there are at least some non-glottochronological pieces of evidence to coordinate against the glottochronological estimate.
    – Khove
    Commented Jun 13, 2021 at 18:35
  • Graeber & Wengrow, in "The Dawn of Everything" have provided lots of new archaeological evidence that challenges the very idea of a rapid "Neolithic Revolution" arguing instead for the Middle East undergoing a transition from foraging & hunting to cultivation over 3,000 years. During this time, as the planet was warming following Last Glacial Maximum, new patterns of river flows emerged that left parts of the land wetter and dryer in different seasons. During this time also people experiment around with planting different grains in the alternating dry/wet soil...
    – abhishek
    Commented Aug 14, 2023 at 15:17
  • ...until they get real good. I've yet to see what evidences Graeber & Wengrow provide for NE Africa specifically, and what was going on there, but it's possible this period of "gathering grains" but no full-on cultivation may actually refer to this (long) time period of experimental proto-farming, and the date ante quem for Proto-Afro-Asiatic might be much later than we thought.
    – abhishek
    Commented Aug 14, 2023 at 15:18

You have raised very strong objections to the proposed dating of Afro-Asiatic. As you say, the only basis for the 14000 figure is “glottochronology”, that is: the unfounded superstition that all languages develop at the same rate and that languages shed vocabulary with the same natural regularity as the breakdown of radioactivity, in short a false analogy between linguistics and nuclear physics.

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.