21 November 2015

A Normally Weird Language

Every week, the digital magazine Aeon publishes several ambitious essays, by competent writers, on culture, philosophy, science, technology and other interesting subjects. One of last week’s authors is John McWhorter, professor of linguistics and American studies at Columbia University; the topic is the English language. The essay is entitled “English is not normal”. Professor McWhorter argues not only that English is genuinely “weird” (anyone who has followed his publications already knows it) but makes a stronger claim that it “really is weirder than pretty much every other language”. Now that is a really weird thing to say, so let’s see how it is argued.

English is not normal
McWhorter begins by discussing English spelling and its caprices (with the reservation that writing is secondary with respect to speech). This is of course due to the conservative character of the spelling system, which has not undergone any major reform since Late Middle English. But English is by no means the only language with such a mismatch between its spoken and written form due to the reluctance of its orthography to catch up with sound change. French, for example, is just as weird. It has plenty of ambiguous spellings with more than one possible pronunciation and alternative spellings for one and the same phoneme in one and the same position. It easily beats English when it comes to mute consonants: vin, vins (verb and noun), vain, vains, vint, vaincs, vainc, vingt are all pronounced /væ̃/. Massive mergers of this kind would surely have caused any normal language to collapse, so French can’t be normal, can it? Irish spelling was even worse before its mid-20th-c. modernisation, and still remains a pretty complicated affair (regular, but you have to master quite a few rules to figure out how to pronounce bhfaighidh). Lhasa Tibetan has lost many consonant both in initial and final clusters, but has retained their spelling representation. And while we are in Asia, isn’t written Chinese even a little weird? Professor McWhorter says that “in countries where English isn’t spoken, there is no such thing as a ‘spelling bee’ competition”. To my knowledge, national spelling competitions are organised in many countries, including Poland. I have finished runner-up in one of them, and I can testify it was tough going. Is Polish a normal language?

The  next claim is that English is not similar enough even to closely related languages to guarantee partial mutual comprehensibility. Well, this depends on what we regard as a “related language”. If, for example, we treat Scots as a close cousin rather than a variety of English, we have to agree that English and Scots are partly comprehensible to each other’s speakers (more so, I presume, than Standard Dutch and High German). English and Frisian are more closely related to each other than either is to the rest of Germanic, but they became separated geographically more than 1500 years ago and, unlike Dutch and German, or Spanish and Portuguese, have not remained in contact or been connected by a continuum of intermediate dialects. If that makes English weird, Greek, Albanian and Armenian are even weirder (not to mention such orphan languages as Japanese, Burushaski or Basque).

According to McWhorter, English is the only Indo-European language without grammatical gender. This sweeping statement is simply false. Let’s begin with the observation that the “classical” three-way distinction (masculine : feminine : neuter) probably did not exist in Proto-Indo-European itself, which only distinguished neuters from non-neuters (a state of affairs thought to be preserved by the extinct Anatolian languages such as Hittite). Once the three-gender system emerged in the rest of the family, it was reduced again in some branches. For example, although Latin had three genders, all the modern Romance language descended from it have only two, having eliminated the neuter. Among the Scandinavian languages, Danish and Swedish have merged the feminine and masculine into one “common” (non-neuter) gender. English has gone one step further. Already at the Early Middle English historical stage all morphological markers of gender were abolished in nouns and adjectives. The only trace of the former three-way system is a “natural gender” distinction in the third person singular of personal pronouns (he : she : it). But even within the Germanic group we find the same development in Afrikaans. If anything is “weird” about gender in English and Afrikaans, it isn’t its loss in nouns, but rather the survival of natural gender in pronouns: having pronominal but no nominal gender is very rare cross-linguistically. As for the rest of the Indo-European family, there is no grammatical gender in modern Persian, Balochi, Ossetic, and several other (though not all) Iranian languages. Armenian (also Indo-European) has no gender either. Both the genderless Iranian languages and Armenian are more consistent than English in their elimination of gender: their personal pronouns are genderless too. Armenian na means ‘he/she/it’; literary Persian has u ‘he/she’ (used only of humans) contrasting with ân ‘it’ (non-human), but the latter has taken place of the former in spoken Persian. As we can see, English is by no means alone even in Indo-European. And since more than 50% languages worldwide have no morphological gender or noun-class system, it is in good company.

The next feature is genuinely weird ­– here I completely agree. No other language known to McWhorter or to me marks the third person singular of present-tense verbs and leaves all the other forms unmarked (the sole exception is the present tense of to be). This is of course due to a historical accident caused by extralinguistic factors – the generalisation of the originally plural polite pronoun ye/you, which led to the disappearance of 2sg. thou/thee together with all the verb forms associated with it (art, wilt, dost, hast, drink(e)st). Nevertheless, it’s strange, though hardly strange enough to justify the claim that English is “deeply peculiar in the structural sense”.

Less convincing is the case for the weirdness of do-support in questions requiring inversion (does she smoke?), in negation (she doesn’t smoke), and in emphatic statements (she does smoke). Professor McWhorter has for a long time argued that the construction is due to Celtic influence and found exclusively in Brittonic Celtic and English. This is doubtful for several reasons. Constructions regarded as precursors of do-support occur sporadically in 14th-c. English, but fully assume their modern functions and begin to spread rapidly after ca. 1500. That’s 1000 years after the initial contact between the Anglo-Saxon and the Brittonic Celts. Why so late? Perhaps the construction existed in informal spoken English and didn’t make it into the written standard until the sixteenth century? Such an explanation could work for Old English, but hardly for the Middle period, from which we have a vast corpus of documents representing different genres, styles, and grammatical registers. There is, furthermore, no evidence of analogous constructions in Celtic pre-dating their début in English, so the direction of influence is uncertain (if it’s influence at all, rather than accidental convergence made likelier by the fact that inversion is used as a syntactic device in both cases). The fact that the Celtic analogue of do-support can also be found in Breton does not prove its great age. Contacts between the Celtic populations of Brittany and Cornwall were regular and intensive until the decline of an independent Duchy of Brittany in the 16th century. Anyway, even if we are dealing with a pattern borrowed from Celtic, English shares it with Welsh, Cornish and Breton, and so can’t be regarded as exceptionally weird in this respect. Again, the claim that such a construction does not occur anywhere else is exaggerated. Do-support analogues have been reported from some Lombard dialects of Northern Italy (the use of the auxiliary fa ‘do’ in questions), and even from Korean (in negation). A related construction (with Old Norse gera ‘prepare, do’) was used in Old Icelandic negation. Even if the English-specific combination of functions is “special”, its components can be found here and there.

The rest of McWhorter’s essay is devoted to the “mongrel vocabulary” of English (with most of it being actually French, Latin or Scandinavian), the richness of synonymy resulting from layers of borrowing, and the impact of Latinate loans on the development of a complex stress system. Though remarkable, these features are hardly unique of even rare. Plenty of languages have been relexified with foreign elements to a comparable degree, and with equally dramatic consequences for their morphology and phonology.

Of course the essay is pop-linguistics, addressed to a general audience, so the author has every right to simplify things for didactic convenience. He justly debunks the all-to-popular idea of English as the “model” language, so ordinary that it can be regarded as a safe testing-ground for linguistic theories (“let’s consider any language – for example, English”). However, in doing so, he errs on the opposite side, trying to make English look more extraordinary than it really is. English does have its structural idiosyncrasies, but so does just about any other human language. Tsakhur (a Northeast Caucasian language) has ‘tourquoise’ as a basic colour term (it’s also weird in having at least about 70 consonant phonemes); Czech is pretty much unique in having a fricative alveolar trill as a phoneme (a sound so rare that the International Phonetic Association has not yet come up a convenient symbol to transcribe it); Hawaiian has [t] and [k] as variants of the same phoneme in its extremely small inventory of consonants; the West !Xóõ language (in Namibia) has 43-111 different clicks (depending on how you analyse the system) in addition to a few dozen other consonants; Winnebago (Siouan) places the main stress on the third mora in longer words, while Macedonian (Slavic) regularly stresses the antipenultimate syllable; in Imonda (in Papua New Guinea) singular and dual nouns are marked with special endings but plurals are expressed as bare stems; Hungarian has 18 noun cases and two basic colour terms for different kinds of ‘red’. Pirahã (in Amazonas, Brazil) has a dozen phonemes (at most), no numerals, and no basic colour terms; the jury is still out on whether it has embedded clauses. On the other hand, it has a rich verb morphology, with an unusually large number od aspects and several shades of evidentiality (expressing the source/reliability of information). There’s a lot of weirdness out there.

The fact is that the total weirdness of a language is not a quantifiable notion. It makes little sense to say that one language is generally weirder than another (as opposed to being weirder in some particular respect). Caprices of history have elevated English to the status of global lingua franca. It doesn’t owe its unique position to any structural features, although the fact that it has an enormous population of speakers is relevant for its current and future evolution. Yes, it has many eccentric features but hardly represents an extreme type of language. “English is not normal”, while a catchy title, is at best a trivial statement that could be true of any language (if you concentrate exclusively on a few selected oddities).



50 comments:

  1. While I of course fundamentally agree with this post, I would pick a bone or two about French spelling and Greek isolation. Almost every French word has a pronunciation fully predictable from its spelling, though the rules are very complicated (oignon is one of the rare exceptions, being pronounced as if spelled ognon). This is a very different story from English, where some 10-15% of words have utterly unpredictable pronunciations except on historical grounds (one being a fine example).

    Greek is not so isolated (within IE) as it is made out to be, because there is a strong convention of calling the other Hellenic languages "dialects of Greek", no matter how mutually unintelligible they are. But Pontic, Cappadocian, and Tsakonian are by world standards separate languages closely related to Greek. Consequently, Hellenic is as much a (first-order) language family as Romance, though all of its languages but one are on their way down the drain, alas.

    ReplyDelete
  2. Let's say that French spelling is weird in a slightly different way. And of course there are quite a few unpredictable pronunciations (including many instances of consonants which are not mute though one would expect them to be). I agree, however, that they are probably much fewer than in English.

    As for Greek being a tiny family rather than a single language, the same point can be made about Albanian (Geg vs.Tosk, not to mention the status of Arbëresh), Armenian (Eastern vs.Western), Japanese (the Ryukyuan languages [plural], Basque (at least five major dialects rather different from each other and from Standard Basque), and even Burushaski (the Yasin dialect ois sometimes considered a separate language). Languages are fuzzy objects.

    ReplyDelete
  3. Most of the rather few French words whose pronunciation cannot be predicted from the spelling are short and common, and most of the rest are proper names. English has exceptions everywhere, while having way fewer spellings with silent letters as you illustrate.

    German is qualitatively more like English, just in a weaksauce way. :-) Tibetan is generally like French on both ends of the word.

    Mongolian-in-the-Mongolian-script appears to be a lot like English: the orthography is historical, and there aren't enough graphemes for all the phonemes (while others have fallen silent much like English gh).

    I've heard claims that northern/Yorkshire English is rather trivial to understand if you speak Low German. Probably exaggerated (if only because Low German doesn't have all those medieval Scandinavian loans!), but I don't know by how much. Hard to test nowadays...

    Different Danish dialects retain 3, 2 or 1 = 0 genders; map on Wikipedia.

    Turquoise... I'd say in modern German, even my admittedly urban dialect, it's "second-order basic" like pink or orange or purple. The name türkis (final stress) is of course a loan, unlike in Tsakhur (I bet), but it's in common use and noticeably more frequent than, say, teal in English (let alone technical terms like mauve, cyan, magenta) – and I'm already counting all the teal deer on the Internet. :-þ

    Czech is pretty much unique in having a fricative alveolar trill as a phoneme (a sound so rare that the International Phonetic Association has not yet come up a convenient symbol to transcribe it)

    It did: ɼ. But later it retired that symbol for, presumably, some reason. It has also retired the symbols for the voiceless implosives (which aren't even in Unicode).

    As for other languages, a few (like Nivkh, because of course...) have the voiceless version; Tarifit (Rif Berber) appears to have the voiced one.

    Winnebago (Siouan) places the main stress on the third mora in longer words

    Does that mean on the 2nd syllable if it's long, and otherwise on the 3rd, the mirror image of Latin?

    ReplyDelete
    Replies
    1. It did: ɼ. But later it retired that symbol for, presumably, some reason.

      I wonder why they can't employ the same quasi-diacritics which are used for voiceless and voiced fricative laterals: [ɬ, ɮ]. It would not be very difficult to combine them graphically with the letter "r" (and promote them to real diacritics).

      Does that mean on the 2nd syllable if it's long, and otherwise on the 3rd, the mirror image of Latin?

      I'm not sure if the second-syllable vowel can ever be long in Winnebago. In trimoraic and longer words, if the first syllable contains a long vowel, the second carries the main stress. If both the first and the second vowels are short, the stress falls on the third. This rule is complicated by the operation of so-called Dorsey's Law (epenthesis with an "echo vowel" in underlying stop+sonorant clusters: /kro/ -> [koro]). In some cases stress rules skip this epenthetic vowel, so the main stress gets shifted to the fourth surface mora, as in hikorohó (the first of the o's is a Dorsey vowel). The whole story would be difficult to explain in one paragraph. The interaction of primary and secondary stress with Dorsey's Law makes Winnebago stress something of a Rube Goldberg system. Very, very weird.

      Delete
    2. "Mongolian-in-the-Mongolian-script appears to be a lot like English: the orthography is historical, and there aren't enough graphemes for all the phonemes (while others have fallen silent much like English gh)."

      To make matters worse, there are too many graphemes for the phonemes in other cases. The velar/uvular series isn't contrastive but they have distinct graphemes, because the script was originally borrowed from a Semitic language in which the distinction was important.

      On top of the fact that it was originally not particularly well-suited for transcribing the language, the traditional script (as used in Inner Mongolia) hasn't changed since Middle Mongol, leaving it quite incapable of indicating modern contrasts these days.

      Delete
    3. When the Mycenaeans adopted the Linear B script, it was totally unsuitable for transcribing Greek (no phonation contrasts, no consonant clusters, no closed syllables, no r/l distinction). They had to write qo-u-ko-ro for *gʷoukolos 'cowherd', ta-to-mo for *statʰmos 'outpost', etc. The Greeks did a much better job the second time, adapting the Phoenician script in a highly creative way.

      Delete
  4. Do-support occasionally shows up in German. It's used rather chaotically, but it's common enough that there's a prescriptivist saying for little children: "tun" tut man nicht, "'doing' is not the done thing". The obvious response is "tun" tut man nicht tun, with do-support on "do" itself.

    I'm surprised you have spelling bees in Poland. I've never heard of any in a German-speaking place – and it's not like people wouldn't make spelling mistakes.

    ReplyDelete
    Replies
    1. The Wikipedia article is unfortunately only in Polish, but you can see my name under 2001 (second place).

      They have similar National Dictations in the Netherlands and Belgium, Lithuania, France, and I suppose many other countries.

      Delete
    2. Ah, dictations, where punctuation etc. also counts. France has those. At a spelling bee the contestants are asked to spell only one word at a time.

      Delete
    3. I actually find English spelling bees easier. Memory is cheap.

      Delete
  5. I've now read the article. Eeny meeny miney moe is so similar to the German* ene mene muh that Vennemann took both as part of the substrate in Proto-Germanic!

    Hebrew easier to learn than Russian? From English as a starting point that might just be possible; I bet it's not true from a German starting point. What little I know of Hebrew grammar looks plenty scary to me.

    * Not southern; I don't know if it can be localized with more precision.

    You've been a bit too unkind on the issue of gender: McWhorter only says that the IE languages in Europe all retain gender, and with the exception of most of Jutland that seems to be correct (for narrower, more traditional definitions of Europe).

    ReplyDelete
    Replies
    1. The wording is somewhat unclear:

      But actually, it’s us who are odd: almost all European languages belong to one family – Indo-European – and of all of them, English is the only one that doesn’t assign genders that way.

      The highlighted them can be read as referring to the languages of (traditionally defined) Europe, the Indo-European family, or the intersection of both classes. The statement is obviously false for the first two interpretations (neither Basque nor any of the many Finno-Ugric languages have nominal or pronominal gender) but correct for the third (which is perhaps what the author meant). Sorry if my interpretation was ungenerous.

      Delete
    2. Eeny meeny miney moe is so similar to the German* ene mene muh that Vennemann took both as part of the substrate in Proto-Germanic!

      I'm not convinced at all it's Celtic -- not the way the sheep-counting rhymes in northern England doubtless are (yan, tan, tethera, pethera, pimp...).

      Delete
    3. I actually think McWhorter confused those.

      Delete
    4. Yeah, hovera, dovera, dick belong to the "yan, tan, tethera" series, not to eena, meena, mina, mo.

      Delete
  6. Middle Dutch also had do-support constructions, though they were not common. Two examples from the early 1300s:

    Artur, coninc goet, doet mi horen. ‘Arthur, good king, do hear me.’
    Der minnen goet, dat alle smerte verdriven doet. ‘The good of love, that does drive away all pain.’

    Even in Modern Dutch, they are still used to express generality:

    Dansen doe ik niet. ‘I don’t dance.’
    Zingen doe je zo. ‘This is how you sing.’ (Lit. ‘So do you sing’)

    And for emphasis:

    Klagen dat hij deed! ‘He was complaining so much!’ (Lit. ‘Complain that he did!’)

    And informal, somewhat jocular speech has the verb as an auxiliary in imperatives, with eens, an adverb which usually means ‘once’, but is often thrown in simply to soften the force:

    Doe eens lachen. ‘Laugh, will you.’ (Lit. ‘Do laugh once.’)

    For further reading, see this Dutch article (with English abstract).

    - Olivier van Renswoude

    ReplyDelete
    Replies
    1. Thanks, Olivier! There's also an interesting PhD thesis by Aaron William Ecay (recently defended at UPenn) on the evolution of English do-support Ecay argues that the phenomenon can be analysed into several subcomponents which are individually common cross-linguistically, though their particular combination happens to be very rare.

      Delete
    2. Cheers, I will give that a read.

      Great blog, by the way!

      - Olivier

      Delete
    3. Your examples for "generality" and "emphasis" are topic-and-comment sentences, something that English is highly averse to even for European standards. They're quite common in German. They are formed by abstracting the verb, as an infinitive, to the first position, leaving its inflection behind in the second position; by embodying the inflection on a dummy verb, we can have our cake and eat it, too.

      And that is exactly what English does in questions and negative imperatives, just the other way around: inflected dummy verb in first position, uninflected content verb behind the subject so that SVO word order can be maintained in (just about) all sentences.

      Delete
  7. What I want to know is, what is his thing about Russian? Practically everything he writes contains some remark about how Russian is the hardest language ever in the history of mankind and he doesn't know how anyone can speak it. Did he have a sadistic/incompetent teacher? You'd think it was Kabardian or something.

    ReplyDelete
    Replies
    1. I also wonder what is so hard about Russian, but John McWhorter thinks it possible that speakers of other Slavic languages are able to cope with it.

      Indeed, in comparison with Polish, Russian is simplicity itself. No personal endings in the past tense, immobilised verb clitics, no vocative unless you adress God, many paradigmatic alternations levelled out (рука : в руке), fewer tongue-twisting consonants clusters, etc. True, ударение can be a problem, but you can't expect a language to be easy in every respect. ;-)

      Delete
  8. Scots also marks the 3rd.sg with -s. Although Scots mostly forms participles to form the present tense, with bare verbs and -s being more gnomic

    ReplyDelete
    Replies
    1. Traditional Scots and Northern English are even weirder than mainstream English. They have the Northern Subject Rule:

      they come : the men comes

      Delete
  9. "all the modern Romance language descended from it have only two, having eliminated the neuter"

    Do you regard the Romanian neuter as 'ambigeneric' then?

    ReplyDelete
    Replies
    1. Yes. Descriptively, some nouns switch their inflectional class with number. Note that adjectives modifying Romanian "neuters" are masculine in the singular and feminine in the plural.

      The same may happen in Italian: l'uovo 'egg' : le uova 'eggs' (Rom. ou : ouă). There are also quite a few Italian masculines with two plurals, usually distinguished semantically, like il dito 'finger' : i diti (count plural) : le dita (collective plural).

      Cf. Lat. locus 'place, spot' : loci '(individual) places' : loca 'region, quarters' -- a pattern which goes back to PIE.

      Delete
  10. OK but in Romanian it's far more systematic. It's not just 'some nouns', it's a large and very productive category.

    And many of them take the "-uri" ending, which is pretty rare among feminine nouns, and looks (to me at any rate) like it results from Latin neuters in -ora like corpus, tempus etc. (with the -i as an additional plural marker analogous to English "children" and "brethren"). But maybe I'm wrong.

    Also, is there any functional difference between saying "there's a large number of nouns that switch their inflectional class with number", and "there's a third gender which shares its inflexions with the masculine singular and feminine plural".

    I fully understand if you don't have time for this excursion from the main point of your post, which is spot on!

    ReplyDelete
    Replies
    1. It's a very good question, and of course opinions are divided. I'm more inclined to see Romanian as a two-gender language because grammatical gender is to me primarily a matter of agreement (rather than just inflections). What one sees in the "neuter" class is masculine agreement in the singular and feminine agreement in the plural. Switching gender with number for those nouns seems to me more parsimonious than the traditional three-gender analysis with only two agreement patterns.

      Membership in the gender-switching class depends to some extent on semantic factors -- no animate nouns belong there. But is this a sufficient reason for distinguishing a separate inanimate/neuter gender? It's quite common for semantics to correlate with inflectional types. English countable nouns with unchanged plurals mostly denote animals -- game or fish (especially in sporting usage), and there's a small but old core of animal names reflecting OE strong neuters (sheep, deer, swine). But it would be wrong to say that in addition to countables and uncountables English has a third category which historically continues the old neuter and is not inflected for number (like uncountable nouns). First, the vast majority of the modern members have no neuter ancestors (not even fish and fowl); most have been drawn into this class by analogy. Secondly, they pattern with ordinary countable nouns in every respect as regards agreement (these trout are swimming like these whales are swinmming).

      Delete
    2. I still find it hard to get past the -uri. It's just so much more marginal a feminine ending than it is a 'neuter' one. It seems a stretch to say that nouns like 'timp' have feminine inflection in the plural. But I do take your point regarding agreement.

      Online (eg Wikipedia article on Romanian grammar) there's some suggestion that the category might be to some extent artificial, a late Latinising programme. These claims are made a lot about Romanian, but I've been out the game too long to know if it's serious scholarship, or disgruntled Hungarian nationalists trying to strip the language of its still prestigious latinicity... or both.

      On another note, I've been rereading your 'water' and 'four' series, which are so compulsive they kept me up till, er, four in the morning. Thanks for that.

      Delete
    3. The grammars of many languages (not only Romance ones) have been shoehorned into the familiar Latin model. Much of the terminology we continue to use shows this traditional Latin-centric bias.

      There's a good (and quite influential) summary of the "two or three genders" controversy by Nicoleta Bateman and Maria Polinsky (2010). You can download it here (the authors argue for a two-gender analysis). Their approach has been used by some Romanian computational linguists to test the (machine) learnability of the two-gender model: Dinu et al. (2012). On the other hand, Ruth Kramer (Morphosyntax of gender, published this year by OUP) defends a compromise between the traditional analysis and the two-gender system (three genders for the purpose of noun classification, two genders in terms of morphologically expressed agreement).

      Im happy if you find my blog series worth rereading :)

      Delete
    4. Thanks, I will definitely check those papers out.

      What I meant about the latinising programme is the frequent claim that Romanian is an 'invented language' - relatinised in the 18thnand 19th century more comprehensively than the other Romance languages.

      Delete
    5. Oh, yeah, limba rumînească > limba română. I'm not sure to what extent the massive import of French and Italian vocabulary has affected the declensional system. I suspect that the evolution of features like the -uri plurals was independent of "Latinisation", and that nouns were attracted into that class based on synchronic formal and semantic factors, not based on their etymology. Of course any highly productive pattern is likely to grow even more productive when thousands of new loans enter the language. For example, nearly all verbs borrowed into English during the past 1500 years (with a few isolated exceptions) have joined the "regular" -ed declension (and many originally strong verbs have migrated there too).

      Delete
  11. Irish also has two terms for "red" (you mentioned this as a unique/rare feature of Hungarian): dearg and rua. I'm not an expert, but I think that rua tends to refer to a more brownish red (rua is used e.g. for red hair), whereas dearg refers more to the standard paint color red.

    ReplyDelete
    Replies
    1. So does Polish, for example. The older word rudy (of course strictly cognate to red and rua, all of them reflecting PIE *[h₁]roudʰ-o-) refers primarily to red(dish) hair or fur, while the dialectal Slavic innovation czerwony is the standard colour term. But the definition of a "basic colour term" (BCT) -- all controversies aside -- requires that its reference should not be restricted to any particular class of objects, or to things that have an inherent colour (as opposed to being artificially dyed or painted).

      Which said, the BCT concept is a little fuzzy, which is only to be expected, given that colour-term inventories are historically evolving systems. Before a term becomes established as "basic", it must pass through a stage when its status is uncertain.

      Delete
    2. P.S. It seems that the situation in Hungarian is unstable, and that rather than dividing the relevant part of the colour space between them as two distinct, specialised BCTs, piros and vörös are competing for BCT status and the older term (vörös) is already losing: Benczes (2014).

      Delete
    3. Sorry, it's Benczes and Tóth-Czifra (2014). The final, published version can be viewed here. I wish Blogger offered an "Edit" facility.

      Delete
    4. Blogger (Google Blogs) doesn't even support the <blockquote> tag...

      PIE *[h₁]roudʰ-o-

      I thought the reconstruction of *h₁ is inevitable because of Greek érythros?

      Delete
    5. Yes, in the root *h₁reudʰ- in general. But in this particular stem the o-grade triggers the so-called Saussure (or Saussure-Hirt) effect. Underlying *HRo and *oRHC sequences lose the laryngeal (or at the very least the laryngeal doesn't produce a vocalised outcome in such derivatives). The [] brackets are intended to mean that the laryngeal is present underlyingly but may have no phonetic realisation.

      Delete
    6. The IE word for 'red' is surely a Wanderwort of Middle East origin (cfr. Sumerian urudu 'copper').

      Delete
    7. Oh, that. That's a strange effect, though; why would specifically *o trigger it...?

      Sumerian urudu 'copper'

      Well.

      Delete
    8. Oh, that. That's a strange effect, though; why would specifically *o trigger it...?

      Presumably it has something to do with PIE *o being a tense vowel (in comparison with *e). Cf. its tendency to lengthen in Indo-Iranian (Brugmann's Law) and the acrostatic ablaut pattern with "strong" *o alternating with "weak" *e.

      Strange as it is, the effect is very securely supported by numerous examples found in languages which preserve laryngeal reflexes well (e.g. Greek, Hittite). Andrew Byrd gave an interesting talk about the "crazy" character of the Saussure effect some time ago. As far as I know, he's prepared a forthcoming publication on it in collaboration with Alexander Hansen, but you can find the preliminary presentation at Academia.edu

      Delete
    9. In my (not humble) opinion, Whittaker's work on Sumerian is on the same league than Cašule's on Burushaski. In short, they fell into the same trap in using the reconstructed "PIE" as a measure of the supposed "IE-ness" of these languages.

      Delete
    10. I've met Gordon at a couple of conferences and had an opportunity to exchange thoughts with him. My opinion is generally positive. To be sure, comparing anything with Sumerian is relatively easy, given the shortness, simple structure and uncertain semantics of Sumerian morphemes. As a result, Sumerian is one of the favourite targets of the linguistic fringe (not to mention complete kooks). But Gordon's approach is careful. He is very open to criticism and has refined his analyses as a result of discussion. The PIE forms he maps onto Sumerian lexemes are impeccable and represent actual reconstructible words, complete with ablaut grades and derivational and inflectional suffixes, so he isn't just another root-equation maniac. Controversial as Euphratic is, I would say Gordon Whittaker has made a pretty solid case for it.

      Delete
    11. I like the presentation! Lezgian becomes more awesome every time I look at it. :-)

      In short, they fell into the same trap in using the reconstructed "PIE" as a measure of the supposed "IE-ness" of these languages.

      If you really think that's the biggest problem with Čašule's publications on Burushaski...

      He is very open to criticism and has refined his analyses as a result of discussion.

      Case in point: I just read a critique from 1999 which takes issue with lots of things that simply aren't in the 2008 version anymore.

      the shortness, simple structure and uncertain semantics of Sumerian morphemes

      Also uncertain phonology, let alone phonetics!

      Delete
  12. I just read the (same?) 2008 paper, "The Case for Euphratic".

    Has he published a systematic account of proposed correspondences between 'Euphratic' words and loans in Sumerian and Akkadian? The correspondences are on the face of it impressive, and as Piotr says, it certainly looks like serious and honest scholarship.

    But in a limited time it's hard to fathom how much leeway he's allowing himself. Given that there are often various different Sumerian versions, plus Akkadian, plus hypothetical alterations inferred from other correspondences, plus occasional variation in reconstructed IE forms, of which different nominal cases are available, plus scope for semantic drift... one would predict an awful lot of coincidental similarities.

    On the other hand, if he's right, it would be among other things spectacular confirmation of the laryngeal theory - knocking the Anatolian languages right out the water!

    ReplyDelete
    Replies
    1. The 2008 paper is a developed version of what Whittaker presented at an IE conference in Copenhagen in 2000 (published 2004 as "Word formation in Euphratic"). He gave another paper at another Copenhagen conference in 2009. It appeared in 2012 as "Euphratic: A phonological sketch", in B. N. Whitehead et al., The sound of Indo-European..., Museum Tusculanum Press, 577-606 (we met at that conference, and I have a paper in the same volume). His theory has evolved, and in the 2012 article he goes to great lengths to lay out a detailed pattern of regular correspondences between IE and Sumerian. Some of the equations seem rather far-fetched to me, and of course a good number of chance resemblances can be expected, given the limitations of the Sumerian writing system and the philological uncertainties. I find it a little amusing that some aspects of Euphratic look vaguely "Italoid". I'm impressed by several individual correspondences rather than the sheer bulk of the material. I'm quite sure much of it is random noise, but it seems as if quite a few real IE loans and Wanderwörter were lurking there.

      Delete
    2. I find it a little amusing that some aspects of Euphratic look vaguely "Italoid".

      I suspect Latin happens to be the ancient IE language Whittaker is most familiar with. :-) But now I'll read the paper!

      Delete
  13. Although I can see the point about English being weird, and in a European countext I think there is truth to it, at a global level there are even weirder languages. For instance let's take Chinese, the only language in the world not to use a phonetic writing system. Now that's weird for real.

    ReplyDelete
    Replies
    1. The Chinese logographic writing system and its derivatives are applied to several other languages, to be sure, and there have been other logographic scripts in the history of language, but you are absolutely right: Chinese is wonderfully weird in many interesting ways :)

      Delete