Biliteral Roots.

The Commenter Known as Y has sent me some intriguing links, with the following introduction: “The basic idea is that hunter-gatherer-type words in Proto-Semitic are based on biconsonantal stems, and that agricultural-type items are triconsonantal. That fits well with the contested idea that triconsonantal stems are a later development. I think it’s a clever and original idea, though the author seems to go a bit overboard with the conclusions.” The new paper is “Statistics of Language Morphology Change: From Biconsonantal Hunters to Triconsonantal Farmers,” by Noam Agmon and Yigal Bloch (PLoS ONE 8 [2013]); its abstract:

Linguistic evolution mirrors cultural evolution, of which one of the most decisive steps was the “agricultural revolution” that occurred 11,000 years ago in W. Asia. Traditional comparative historical linguistics becomes inaccurate for time depths greater than, say, 10 kyr. Therefore it is difficult to determine whether decisive events in human prehistory have had an observable impact on human language. Here we supplement the traditional methodology with independent statistical measures showing that following the transition to agriculture, languages of W. Asia underwent a transition from biconsonantal (2c) to triconsonantal (3c) morphology. Two independent proofs for this are provided. Firstly the reconstructed Proto-Semitic fire and hunting lexicons are predominantly 2c, whereas the farming lexicon is almost exclusively 3c in structure. Secondly, while Biblical verbs show the usual Zipf exponent of about 1, their 2c subset exhibits a larger exponent. After the 2c > 3c transition, this could arise from a faster decay in the frequency of use of the less common 2c verbs. Using an established frequency-dependent word replacement rate, we calculate that the observed increase in the Zipf exponent has occurred over the 7,500 years predating Biblical Hebrew namely, starting with the transition to agriculture.

An earlier paper by Agmon alone is “Materials and Language: Pre-Semitic Root Structure Change Concomitant with Transition to Agriculture” (Brill’s Annual of Afroasiatic Languages and Linguistics 2 [2010]); it was preceded in the journal by “An Introductory Note to Noam Agmon’s ‘Materials and Language’ with Special Attention to the Issue of Biliteral Roots,” by Jean Lowenstamm, whose abstract begins “Biliteral roots have been, and still are controversial.” (All links are pdf.) I look forward to seeing what my readers who have ideas about Semitic biliteral roots have to say.


  1. David Eddyshaw says

    “Linguistic evolution mirrors cultural evolution”

    … let me stop you right there …

    Seems wide open to cherry-picking to fit the thesis.

    The usual Biblical Hebrew word for “wine”, for example, is not 3-lit “chemer” – it’s “yayin”, which would be 2-lit according to their scheme.

    “Field” in BH is normally “sadeh”, whichis 2-lit by their criteria.

    It doesn’t inspire confidence that they use Orel’s recently-LH-mentioned Afroasiatic dictionary as a source, to say nothing of the Nostratic.

    The stuff about Zipf’s law strikes me as hand-waving. I don’t think anybody understands why Zipf’s law works for languages in the first place really, and to pray it in aid of a proof strikes me as chutzpah, particularly when applied to a text like the Hebrew Bible full of hapax legomena, long passages of poetry full of rare and sometimes obscure words and a great mixture of genres.

  2. Heh. Yes, I too felt the aroma of chutzpah here and there.

  3. David Eddyshaw says

    “Village” may be 3-lit, but “city” – “ir” is 2 lit (sorry, I really ought to find out how to type Hebrew on this machine).

    Perhaps our pre-agricultural forbears had cities but no villages.

    You could make a party game out of this. OK, maybe only if you chose the guests very carefully.

    It would be in no way surprising to find that 2-lit words were, in general, a more fundamental part of the vocabulary in some sense, just as shorter words tend to be in all languages.

    So one crucial thing that needs to be established is whether the agricultural-era words can be systematically picked out by some method not open to unconscious bias in the selection. It’s hard, admittedly, to imagine a way of doing this, but if you can’t, any supposedly positive result is just likely to be due to the incontestable fact that reaping corn is a less basic human activity than sleeping or eating. [Though I worry about that aleph in the word for eating. Three radicals in my book. Did stone-age man eat?]

    The agricultural revolution presumably must indeed have had an impact on the languages of the first farmers and caused the introduction of some new, less-core, vocabulary. But this long antedates any written langages, and so such effects would need to be sought in prehistory and reconstructed protolanguages. This is just where the reliance on flaky sources like Orel and the like is worrisome. A particular problem is going to be the systematic failure to distinguish borrowing from common descent that plagues such works.

  4. marie-lucie says

    Within the history of languages, it is well established that societal upheavals and population movements and mixes are important factors in language change. Changes can affect all the constituents of language, ranging from vocabulary change (replacement through borrowing, formation of new words, semantic widening or narrowing), pronunciation changes, syntactic change (especially in cases of widespread bilingualism) and, least likely, morphological change (adoption of morphemes or types of word-formation from other languages). For instance, the French presence in England after the Norman conquest had considerable influence on the vocabulary, some influence on syntax, but did not affect verb morphology, which remained typically Germanic. For the Hebrew examples, the adoption of agriculture, having major influence on subsistence, mobility, land tenure, and associated changes in social structure, could not fail to have repercussions on the languages spoken in the area, but to associate a major change in morphological structure with specific changes in sociocultural conditions is going a bit far. I have not been able to read the articles, so I am not sure whether my conjecture makes sense in terms of the actual history of those languages, but going from biliteral roots (some of which still exist) to triliteral ones cannot have been a sudden jump. It is more likely that there were some additions to the CC root, at first optional, hence CC+C stems alternating with bare-root CC ones in the same words; later the extra C was frozen in place, and morphological structure was somewhat reorganized, treating the new CCC sequences as unanalyzable roots. This sort of development could occur spontaneously over a long period, whether or not there were significant changes in the people’s way of life. For instance, it one Semitic people adopted agriculture, would its still nomadic neighbours, living far from the fertile area, have retained the CC root much longer?

  5. As I recall, Athanasius Kircher claimed that the word for ‘God’ is four letters in all languages. It works for a lot of familiar languages (Fr. dieu, Sp. dios, G. Gott, L. deus, Gr. theos = 4 Greek letters), but he was forced to pretend that the English word for ‘God’ is ‘Good’.

  6. Their “broad definition of 2c roots” seems very strange to me: they justify their inclusion of I-n, I-w, and I-y roots (i.e., פ״נ and פ״י roots) by saying, “These added consonants may represent early affixes, later perceived as radicals”; but they give no evidence for that, and of course the salient common factor here is that those roots lose those consonants in certain forms (lipol, hipil, nolad, holid, etc.), making them appear biliteral. (Some linguists do regard e.g. hikir as biliteral, but that’s an account of the verbs’ current behavior, not an explanation of their history.) At the risk of assuming the worst . . . it’s almost as though they initially included these verbs out of naïveté, and then they learned the etymology but wanted to keep them anyway, so invented an ad hoc explanation to justify it.

  7. I worry about that aleph in the word for eating. Three radicals in my book. Did stone-age man eat?

    Maybe he just didn’t have a word for it. You don’t have to talk about eating in order to eat. It would suffice to mutter to oneself something like: “Me Tarzan you game”.

  8. There may well have been no word for “proposition” either, although it was probably practiced. It would have sufficed to say: “Me Tarzan. You game ?”

  9. “The usual Biblical Hebrew word for “wine”, for example, is not 3-lit “chemer” – it’s “yayin”, which would be 2-lit according to their scheme.”

    yayin is a Mediterranean Wanderwort, borrowed after the rise of agriculture, and not something indigenous to Semitic, so whether it is 2-lit or 3-lit, it is irrelevant to a discussion of Proto-Semitic roots.

  10. “Me Tarzan you game”.

    I believe the first dish that Jane prepared for Tarzan was finch and chimps.

  11. I’m not sure how relevant it is to this discussion, but even a rank amateur like myself can run his eye down a page of a dictionary of Hebrew, Arabic or Egyptian and see that 3c roots beginning with the same two consonants tend to have obviously related meanings, as if the 3c roots are built up from 2c ones. Whether this has anything to do with the coming of agriculturee is of course a whole nother question.

  12. As more communication about more things is “needed”, the stock of 2c combinations will tend to run out. Then you can move up to 3c combinations, or else reconsider whether you need to communicate so much.

    A “transition to agriculture” would therefore make it possible to stick with 2c. When you’re plowing the fields, there’s no time to gabble about a zillion other things, and nobody to talk to but the oxen. Farmers don’t need that new-fangled 3c stuff.

  13. Perhaps some farmers are not laconic because they are farmers, but became farmers because they are laconic. I know that my interest in nC activities like philosophy grew out of frustration at the 3C vapidities of everyday conversation.

  14. All the papers linked to by Hat are beyond my competence to judge, though I wonder about some of the assertions of what constitutes a biliteral root. That said, the papers are fascinating. I suspect Agmon and Bloch are onto something: the development of agriculture may indeed have been the trigger that caused the expansion from a biliteral to a triliteral root system.

    But as Rodger C notes, it’s pretty easy to see what appears to be an expression of this phenomenon running an eye down the page of a Semitic dictionary.

    Example from Hebrew: פה peh means mouth. (It also means ‘here’, and I’m not about to go there.) By Agmon and Bloch this might even be a uniliteral root. In any case, a mouth is an opening, an entry and exit point.

    פהק pahak means yawn. פלט palat means expel.

    The פ-צ p-tz combination seems to have been very productive: פצח patzakh means cracked. פצל patzal means split up or subdivided. פצץ patzatz means exploded or blown up. פצע petza’ is wound, injury.

    פרץ paratz means breached. פתח patakh means opened. פיכה pica means gush. פעי pi’ee means bleat. פער pi’ar means gape, gap.

    I stress that these examples are only what appears to be an expression of the phenomenon.

  15. David,
    “The usual Biblical Hebrew word for “wine”, for example, is not 3-lit “chemer” – it’s “yayin”, which would be 2-lit according to their scheme.”

    As Christopher says, this is a wanderword anyway, but my two cents is that there is no reason to think that wine-making only got going with agriculture. Wine-making requires storage and that requires that you be sedentary, but there is all kinds of evidence from Northe America for instance that lots and lots of pre-agricultural societies were sedentary if they were on good enough land.

    The native range of grapes is well north and east of any possible Proto-Semitic homeland, but as Christopher points out, that’s irrelevant anyway.

    Flash – question for you M-L. The Washo, who were mostly nomadic, would come down from the Great Basin and spend the winter in Nisenan towns in the Sierra foothills, and not to gather food. They didn’t have to gather anything – their huge loads of pinon pinenuts would jazz up a winter’s worth of acorn mush and would make them welcome anywhere.

    Here’s the question – given the wide divergences within Maiduan, what chance is there that the Nisenan population was originally Washo-speaking and that the Washo were just continuing the pattern of visiting relatives in the foothills even after they had language shifted? To your knowledge has anyone ever compared Nisenan and Washo for similarities despite the fact that they are obviously not genetically related?

    Just wondering.

  16. marie-lucie says


    – wine: indeed, people don’t have to be practicing agriculture to have access to wine, if their neighbours and trading partners are making it and storing it in easily transportable containers.

    – Washo and Nisenan: I don’t have much information about Nisenan except that it is Maiduan, but your hypothesis makes sense. In general terms, some of the Penutian languages have borrowed a lot of vocabulary from their neighbours (and sometimes some grammatical items), and no doubt some of the structural differences (especially grammatical simplification) can be attributed to language shift in favour of Penutian. If Washo people were the ones who migrated to spend winters in a warmer climate, it would make sense that they were the ones to shift their language to that of the more favoured location.

    If you look at the map of languages compared to a good physical map, Penutian languages are largely concentrated along rivers emptying into the Pacific. This suggests a migration (or series of migrations) along the coast rather than through the interior, gaining speakers as local populations of the river valleys shifted to the languages of newcomers, meanwhile leaving their mark on the new languages. It used to be thought that the Penutian languages of California were the “core” of the group. According to my research, there is a “core”, but it does not consist in a group of existing languages but a group of morphological features, some of them half-buried in strange irregularities in a series of attested languages.

    I can’t answer your question about a comparative study. For many years the Maidu and Nisenan specialist was Bill Shipley, who died a few years ago. He may have left some documents in an archive or such. The current Nisenan specialist is Eatough.

  17. marie-lucie says

    p.s. If Washo people were the ones who migrated to spend winters in a warmer climate, it would make sense that they were the ones to shift their language to that of the more favoured location.

    Indeed, that warmer location may have been part of their original territory before Penutian speakers settled in it year-round. The newcomers may have been more skilled in using the resources of a river (such as catching and preserving fish), and if the two peoples coexisted on the same territory it suggests that they did not interfere with each other’s use of different resources.

  18. David Eddyshaw says

    @Christopher Culver:

    “Yayin” certainly is a Wanderwort, but it’s old enough in Canaan to have done the Canaanite initial w -> y thing (in fact that’s even in Ugaritic) and has correlates in Arabic and Ethiopic which you would immediately take for straightforward cognates if there weren’t extraneous reasons for thinking that this was borrowing.

    Although I just picked ‘yayin’ off the top of my head, that’s in fact, though, much what I was driving at in saying that there is likely to be a problem in using sources like Orel which are not much good at distinguishing common descent from borrowing. That’s going to be a big issue in exactly this context, because agricultural-revolution originated terms are surely extremely likely to have become Wanderwoerter – like “wine” itself. So there is surely a real risk that a word which isn’t actually Semitic at all originally will end up acquiring a quite plausible but entirely spurious Proto-Semitic forebear, along with its three radicals.

    “Heykhal” – “palace” is definitely Sumerian, for example, but if you didn’t know that you might end up reconstructing a Proto-Semitic haykal- from Hebrew and Arabic etc and concluding that the speakers or Proto-Semitic already had palaces in 6,000 BC or whenever you date it to.

    I suppose you could argue, though, that that doesn’t vitiate the thesis, in the sense that the various Semitic languages would still have got more matching, apparently cognate, 3-lit roots as a result of the agricultural revolution. But it wouldn’t be to do with the effects on Proto-Semitic itself, or have much bearing on the question of triliteralism in Semitic roots and when it became firmly established enough that 2-lit roots got forced into the same mould.

  19. David Eddyshaw says

    @Roger C:

    This business about 3-lit roots seemingly deriving from extension of 2-lit roots has an honourable pedigree to it.
    There are a lot of completely uncontroversial cases where two-consonant root has got extended to three by doubling one of them, or by adding w or y at the beginning or end or in the middle. There are variations like this even in the same language, eg Hebrew tov “is good” (root twb) makes imperfective yitav (root ytb.)

    It doesn’t always seem quite so clear what’s going on when the “additional” letter isn’t one that figures in normal derivational and flexional Semitic morphology. Blau’s “Phonology and Morphology of Biblical Hebrew” is quite scathing about this sort of thing (p189) and claims it had its heyday in the latter part of the 19th century, after quoting one plausible looking case which can actually be shown to be mere coincidence by comparative work. I don’t know to what extent this is his personal take, or whether this is indeed the more recent consensus. He does quote some heavy-duty names as agreeing with him.

  20. David Eddyshaw says

    “Completely uncontroversial” is just asking for trouble, I guess. As Ran points out above, in most cases the y w don’t in fact appear as consonants in the surface forms, although the standard grammars talk about peh-yod verbs, ayin-vav verbs and so forth. There seem to have been a lot of analogical changes, quite possibly in both directions, in fitting 2-lit and 3-lit roots together into a sort of unified framework. (The Ungnad-Matous Akkadian grammar actually associates a characteristic vowel with each 3-lit verb root, and interprets the long u or i of ayin-vav and ayin-yod verbs as representing that vowel in the case of 2-lit verbs. But Akkadian is peculiar in the thoroughness with which it has dropped lots of perfectly good Semitic consonants even by the beginning of the second millennium.)

    Nevertheless the basic point is still valid – a root with two consonants may get adapted to into a 3-lit type structure in more than one way, even in the same language.

    Moreover the ways this happens differ in the details from language to language, so the adaptation must have still been a work in progress in Proto-Semitic.

  21. I think there’s good evidence for considering at least some nCC roots as extensions of earlier biliterals, since n- is a productive verbal prefix in attested Semitic languages, and there are a number of cases where the semantics are close, at least in Hebrew:

    נבל nbl “wither, fade, waste away” : בלה blh* “waste away”
    נדח ndḥ “thrust out, expel” : דחה dḥh*, דחף dḥp, דחק dḥq, all meaning “push, press, thrust”
    נהם nhm “growl, grumble” : המה hmh* “hum, growl, murmur”
    נחש naḥaš “snake” : חש ḥš “be quick”

    And for Nostraticists,
    נטל nṭl “lift, take, pick up” : PIE *telh2- “lift”

    * There isn’t really a final h in these roots except orthographically; the final consonant is null in most forms, t in the infinitive.

  22. There are lots of comments on individual reconstructions in the supplementary material to the article, here.

  23. I think there are some evidences to biliteral roots in semitic languages:

    1) consonants rarely doubled in words but the second and third very often doubled.

    2) triliteral words are harder to pronounce with one vowel patterns (qatl, qitl, qutl) but perfectly fit in biliteral words.

    3) a true triliteral independent roots expected to show some vowel “gravity” meaning in the third consonant, but in fact as regard to the vowel patterns the third consonant is always marginal.

    4) tetraliteral roots are usually a doubled biliteral root or doubled third consonant. There are few roots that integrated a middle consonant (ר and ל usually) but the added consonant always is a non meaning bearing in the vowel pattern.

Speak Your Mind