Post-Neolithic Fricatives.

People keep pointing me to this story (thanks, Bonnie, John, Frank, and anyone I’m forgetting!), so I’m posting it, despite my inherent skepticism. There’s a new Science article, “Human sound systems are shaped by post-Neolithic changes in bite configuration” by D. E. Blasi, S. Moran, S. R. Moisik, P. Widmer, D. Dediu, and B. Bickel, that supports an old conjecture of Hockett’s; here’s the abstract, which begins:

Linguistic diversity, now and in the past, is widely regarded to be independent of biological changes that took place after the emergence of Homo sapiens. We show converging evidence from paleoanthropology, speech biomechanics, ethnography, and historical linguistics that labiodental sounds (such as “f” and “v”) were innovated after the Neolithic. Changes in diet attributable to food-processing technologies modified the human bite from an edge-to-edge configuration to one that preserves adolescent overbite and overjet into adulthood. This change favored the emergence and maintenance of labiodentals. Our findings suggest that language is shaped not only by the contingencies of its history, but also by culturally induced changes in human biology.

It’s been written up in the Guardian and the NY Times (and doubtless elsewhere), and Mark Liberman has a sensible response at the Log:

I agree with Ray Jackendoff that the idea is “is interesting but not earthshaking” — the contribution is not so much a partial explanation for the distribution of labiodentals, because basically who cares, but rather some support for the general concept that physical population differences in principle might sometimes affect language structure. Which again is obviously true in principle, but it’s not clear how often it applies in practice. This result would move the needle from “maybe never” to “apparently once in a while”.

The usual line of reasoning is the opposite, that vocal tract anatomy has (co-)evolved over the eons to serve the needs of speech communication. (See e.g. the section on “Vocal tract changes in hominid evolution” in my lecture notes for ling001.) That seems pretty well supported, though as with functional-evolutionary explanations for anything, there are disagreements.

I have little interest in unprovable origin stories, but boy, people sure do like to speculate.


  1. David Eddyshaw says:

    The Guardian article is (I am sorry to say) particularly stupid. The “science” editor seems either to think that the English word “fuck” emerged four thousand years ago, or (more likely, alas) thinks you can write any old bollocks if the science in question is merely linguistics.

    NYT has a vastly better article, helped by the absence of juvenile sniggering.

  2. Yes, I rolled my eyes energetically at that opening.

  3. David Marjanović says:

    This is not about fricatives, it’s about labiodental consonants (almost all of which are fricatives or approximants for unrelated reasons). I recommend reading the surprisingly long paper and the relevant parts of the supplementary information, both freely accessible from links in the LL comment thread.

  4. John Cowan says:

    An obvious (though not definitive) counterexample is Old Chinese, which lacked /f/ despite having been spoken by agriculturalists since Shang times at least. We don’t see /f/ until Early Modern Chinese, spoken in the Northern and Southern period almost 2000 years later.

    But indeed, where would journalism be without a good dose of juvenile snickering?

  5. Trond Engen says:

    I read a comment to the effect that obviously, once agriculture was developed, being able to say waffle became an evolutionary advantage.

  6. David Eddyshaw says:

    But indeed, where would journalism be without a good dose of juvenile snickering?

    Pravda, I suppose.

  7. Trond Engen says:

    David M.: labiodental consonants (almost all of which are fricatives or approximants for unrelated reasons)

    Not sure about ‘unrelated’. I think labial fricatives often are labiodentals because that’s the simplest way to make a narrow opening between the lips.

    With a modern jaw, I mean. Preagricultural language would have had bilabial fricatives instead.

  8. dainichi says:


    Labial + dental ~> fricative / approximant
    Labial + fricative ~> (labio)dental

    are different propositions.

    As for the first one, the only labiodental stop Wikipedia lists is an allophone of /p/ which assimilates to a following /f/. I’m guessing they are rare globally because some people have gaps between their teeth, but maybe there could be an acoustic reason too?

    As for the second, bilabial fricatives aren’t exactly rare globally, but possibly slightly rarer than labiodental fricatives.

    I admit I haven’t read the whole paper, but are labiodentals really that hard to produce with edge-to-edge bites? I had a bit of an edge-to-edge bite before I got braces, but I don’t remember ever having problems with them. I mean, compared to feats like clicks, this feels like easy stuff to me, but what do I know…

  9. Russian also lacked /f/ despite being spoken by agriculturalists ever since PIE.

  10. In grad school during the 1970s I heard Charles Hockett give a talk on “F-stops and cereal diet” at the University of Hawai‘i. I had been cutting my teeth on Yapese, a language with not just f-stops, but a whole lot of glottalized consonants. The Yapese staple had been very fibrous and chewy swamp taro (Cyrtosperma) for many centuries–you could almost hear it hit bottom when you swallowed a piece. Most Yapese also had a habit of constantly chewing betel nut, husk and all. Meanwhile most Filipinos, despite eating rice for at least a millennium, seemed to have trouble pronouncing /f/, typically substituting /p/. Hockett received a respectful hearing, but nobody was buying his thesis.

  11. SFReader says:

    The Philippines is the only country in the world whose inhabitants can’t pronounce correctly the name of their own country.

  12. What about Palestine?

  13. SFReader says:

    That’s good one.

    And a century ago, an average Russian would have pronounced Russian Federation as “Raseiskaya Khvederatsiya”

  14. SFReader says:

    Tajiks tend to pronounce /p/ instead of /f/ and /s/ instead of /ts/ in foreign words.

    So when a Tajik gastarbaiter tries to pronounce Russian Federation, the second word becomes Pederasia

  15. A visual joke:
    A woman tries to explain where she lives:
    Ya zhivu v Pergane (=Fergane). Vot takoi P (with arms akimbo)!

  16. In the languages of Western North America /f/ is very rare. Only three or four of the languages of California and Oregon have it (SE and NE Pomo, Karuk, and maybe some Kalapuya).

    It’s as if linguistic founder effects were what mattered most.

  17. “you can write any old bollocks if the science in question is merely linguistics” — I think that all science reporting in even “quaity” UK newspapers is liable to be framed as “what wackiness will those boffins come up with next?” If a new finding/theory/invention is not likely to have an immediate relevance to the General Reader, the journo will shoehorn in a spurious hook, e.g. comparison to some scifi trope, or jocular absurd extrapolation. The idea that General Readers mght actually be interested in the science itself seems to have been discounted.

  18. David Marjanović says:

    With a modern jaw, I mean. Preagricultural language would have had bilabial fricatives instead.

    See, that is the point of the paper.

    the only labiodental stop Wikipedia lists is an allophone of /p/ which assimilates to a following /f/

    Yeah, I should have deleted it, because the labiodental affricate is already listed.

    Russian also lacked /f/ despite being spoken by agriculturalists ever since PIE.

    If it had a stable [ɸ], that would be a counterexample.

    Again: just read the paper. There’s a link to a free pdf in the Language Log thread.

  19. David Eddyshaw says:

    The historical angle on this is not subject to proof or disproof, because orthodox comparative methods can’t take you back before the invention of agriculture for the groups which now have it, so the validity or otherwise of the thesis really turns on synchronic comparison of the languages of hunter-gatherers versus everyone else.

    Australia is going to skew the sample hugely, given that, of good descriptions of languages of hunter-gatherers, a very high proportion must be of Australian languages. Even Amazonians like the Jarawara (who have /f/ but not /p/, incidentally, both proto-Arawa /p/ and /pʰ/ having become /f/) have slash-and-burn.

    I’m not aware of any evidence that when hunter-gatherers end up losing their own languages (as they practically all do) they have particular problems with pronouncing [f] or [v] in whatever new language they end up with. Perhaps that’s the benefit of feeding their children Weetabix.

  20. David Eddyshaw says:

    Having now read the paper (good call, David M), I see that the meat of it really is in the synchronic stuff; the diachronic bits involve a fair bit of special pleading to make them apply to [f] but not equally to [θ] or [x], as does their discussion of Australian languages, which do indeed play a starring role, along with “Khoisan.”

    As both the Australian languages and the pre-Bantu southern African languages famously have unusual phonological systems in a great many respects, and are evidently phonological Sprachbünde, it is methodologically suspect to count them all separately statistically. I see the paper has a bad conscience about this to the extent that they did rerun the stats without Australia (duly weakening the result), but the fact remains that they have probably got their priors very wrong.

  21. It’s as if linguistic founder effects were what mattered most

    Perhaps, but in that case, where are all the proto-languages that did have *f? None in PIE, none in PU, none in PTungusic, none in PSemitic, none in PAustronesian, none in PBantu… to name a few families where some languages have /f/ (and often have been retaining it quite well on the subgroup level, e.g. Germanic, Romance, Iranian, Ethiosemitic, Polynesian; showing that /f/ is not an intrinsically unstable transient sound the way e.g. /θ/ is). If there really was a physiological innovation that enabled labiodentals, that won’t lead to all affected populations immediately adopting them, just to a slow creep towards a new equilibrium (insofar as we can speak of such a thing) as language change runs its course.

    One question also seems to be going unasked. If an overbite enables labiodentals, we expect it to inversely disable dentilabials. Indeed today this place of articulation is only found in the extIPA for “disordered speech” (I’ve never heard of an example of them as normative phonemes). Acoustically and, perhaps, phonologically they’re pretty much just a mirror image of labiodentals though.

  22. David Eddyshaw says:

    where are all the proto-languages that did have *f

    Good point. Proto-Oti-Volta certainly did, but the time depth there is probably a good bit less than PIE* and in any case you can reconstruct the word for “millet beer” (for instance) right the way back to the protolanguage … no hunter-gatherers there.

    *Tarim-Liffey is nowadays the preferred term.

  23. John Cowan says:

    Russian also lacked /f/

    Indeed, even Late Common Slavic did not have it (the period, ending around 1000, when the Slavic languages were separated but it was still easy for innovations to pass from one to another).

  24. /v/ or at least [ʋ] appears quite reconstructible for Proto-Slavic though (goes well in hand with the elimination of all u-diphthongs i.e. coda *w).

  25. SFReader says:


    Maritsa-Kamchatka languages

  26. January First-of-May says:

    Russian also lacked /f/

    Which led to… rather weird… results in Byzantine Greek loans with /θ/>/f/ – the Middle Ukrainian reflex of Theodoros is Hvedir.

  27. @Pystynen, the lack of /f/’s in reconstructed languages is a very interesting observation. Here are a few possible explanations:
    1. Something about reconstruction methodology tends to disprefer reconstructing f’s. I don’t know the details of reconstruction of any of these protolanguages.
    2. /f/ either stays put, or vanishes altogether, by way of *f>h>∅ or the like, leaving no traces in the attested record.
    3. /f/ used to be a rare sound, but became common in Eurasia and Africa through language contact in the last few millenia, making its abundance an historical accident.
    4. The sample (IE, Uralic, Bantu) is unrepresentative of reconstructed proto-languages.

    I made all of these up. I don’t know if any of them hold water.

  28. David Marjanović says:

    where are all the proto-languages that did have *f?

    Proto-Caucasian, mwahah.

    However, if you accept that reconstruction* in the first place (let me be the first to point out some pretty big gaps in the data), I’m not aware of reasons to think it was specifically [f] rather than [ɸ]. It was lost pretty often, is reconstructed as being rather rare (more so than *pʰ and *p’, IIRC), and supposedly came from Proto-Dené-Caucasian consonant clusters like *xw and *xŋw.

    Of course, Proto-Caucasian would most likely have been spoken by farmers. It’s not supposed to be much older than PIE, and agriculture has a long history in the region.

    * 200-page “preface” of Nikolayev & Starostin (1994, though really mostly much older).

  29. If labiodentals are easier for modern humans to pronounce than bilabials, why don’t babies go “va-va-va”, and why isn’t “fafa” near-universal baby talk?

  30. David Marjanović says:

    Because plosives are easier than fricatives – they require more power, but less precision.

  31. Savalonôs says:

    I wondered how many reconstructed proto-langs actually had labiodental sounds that are invisible in the reconstruction. Who knows, maybe some or all instances of h₁ in PIE were really /f/. Or maybe there were instances of /f/ in other locations that have lenited to ∅ without traces in all daughter langs. Are labiodental fricatives especially labile in terms of leniting to /h/ or ∅? That is my vague impression of /f/ at any rate.

  32. Trond Engen says:

    David M.: just read the paper

    Science may support my intuitions or go against them. Either way it’s a waste of time.

    Ok. I read the paper. It supported my intuitions. Mostly.


    This is not a cultural change correlated to the development of agriculture but a physiological effect of a change in diet. To find support or falsification in modern languages, the interesting question is not exactly whether traditional hunter-gatherers have bilabials or labiodentals, but whether or not the existence and/or point of articulation of labial fricative(s) in a language is correlated to the rate of overbite among its speakers. Which may or may not be correlated to their lifestyle.

    The absence of *f in proto-languages is interesting, but I was going to say that it doesn’t have anything to do with the development of labiodentals. The introduction of the overbite didn’t introduce new phonemes. The bilabial fricative became more laborious and the labiodental fricative less laborious, which moved the point of articulation of the fricative(s) in the labial series — no doubt to ridicule from old flat-toothers who thought overbiters sounded like children. But then I realized that the point people have been trying to make in this thread is that pre-labiodental f [ɸ] may have been inherently unstable in the way that h and especially þ seem to be. In which case we should reconstruct proto-speakers with edge-to-edge bites. Which is what the paper tries to do with Indo-European.

  33. David Marjanović says:

    Are labiodental fricatives especially labile in terms of leniting to /h/ or ∅?

    Maybe; but I suspect this is instead true of bilabial [ɸ].

    I’ve wondered myself if some or all of PIE *h₁ was actually **[pʰ] > **[ɸ] > *[h], because that just might explain why *b was so extremely rare or didn’t exist at all.

  34. Guillaume Jacques, who is always worth reading, weighs in on his blog. He discusses one language family, Muskogean, some of whose members have an /f/ phoneme. Haas did not reconstruct *f in her Proto-Muskogean. Jacques suggests that the lack of *f in protolanguages might say something about the bias of historical linguists, and not necessarily about the rarity of that phoneme in older times.

  35. David Marjanović says:

    The supplementary information of Blasi et al. contains probabilistic reconstructions of a lot of PIE phonemes to check whether they were labiodental in PIE. They all come out negative, though barely so for *w (if it is, of course, assumed that all sound changes can go in both directions with equal ease). That might be interesting to do for Muskogean.


  1. […] Hat reports, with skepticism, on the idea of “f” and “v” as sounds being products of […]

Speak Your Mind