Two from Foreman.

I wish Alex Foreman were blogging rather than casting his seeds on the stony soil of Facebook, but I will rescue a couple of thought-provoking recent posts from there:

1) You can see phonemic vowel length disappearing in real time over the course of the 20th century in Korean. It’s really well documented. It’s fascinating, and suggests a lot about what analogous processes in languages like Latin or Greek must have looked like. The loss of the length-system is preceded by a period of length instability, during which length is strongly subject to frequency effects and analogy, and all kinds of other interesting stuff happens. Speakers may retain the ability to produce length contrasts consciously for at least a generation after they stop producing them under normal conditions.
Like other sound changes, it did not happen all at once.

2) My favorite example of how versification does not simply flow inexorably from the aurally perceptible, innate rhythmic properties of language is Hindi vs. Urdu.
Urdu and Hindi are, phonologically, basically the same language, and yet their versification systems are really quite different. Hindi meters are basically moraic, whereas Urdu versification draws on Persian metrics and so is much less so.
The result is that Urdu speakers have a hard time sensing the rhythm of a Hindi poem, or figuring out how to scan its lines.
To the “Urdu ear” unaccustomed to Hindi poetry, the verses are often bizarrely long, and the syllable pattern can feel alarmingly erratic. The Urdu listener must get the “feel” of it, learn to hear all over again. Even though the phonology is identical.
It is of course possible to use Hindi meters in Urdu and vice versa. Experimenters have done so. But the difference in rhythmic sensibility is real and stark.

(I have quoted the entirety of each post, so in case they are private on FB you aren’t missing anything.)

Comments

  1. A semester of Hindi in grad school gave me only the basics, but the introduction to Christopher Shackle and Rupert Snell’s “Hindi and Urdu Since 1800: a Common Reader” (you can find it on hindiurduflagship.org, I assume with permission?) made clearer why I could understand the audio CD for “Teach Yourself Urdu” just about as well as the audio for my beginning Hindi class. It’s fascinating to hear about this deeper look into the relationship.

  2. … phonemic vowel length disappearing …

    I’m always bemused by these claims some phonemic (sub-)system is “disappearing” or a language is ‘losing’ some distinction. As if there’s a sort of language entropy/all languages will end up using a handful of phonemes.

    But the Great English Vowel Shift ended up with nearly as many distinct vowels as it started with; Old -> Middle Chinese ‘loss’ of final consonant (cluster)s ended up with tones — 11 or maybe 8 of in the case of Cantonese (depending which analysis you follow), including length distinctions. Proto-Germanic to English ‘losing’ inflections was compensated by greater complexity in syntax, etc, etc.

    analogous processes in languages like Latin or Greek

    In Korean, what’s compensating for the loss of distinctions available from vowel length? Foreman’s giving only half the story without that. (For example is there greater use of particles? or of ‘function words’ that were largely omitted previously?) And so on what basis is Foreman claiming Greek/Latin processes are “analogous”?

  3. PlasticPaddy says

    Give it time, say 3 generations. Maybe Foreman is using ‘analogous’ to mean ‘similar’ or is restricting the analogy. For you, can two things have analogous forms but separate origins/functions/effects (or vice-versa, as Sir Boyle Roche would say; German Föhn = (1) steady warm dry wind and (2) hair-dryer.

  4. Stu Clayton says

    Also both: German Stöhn = (1) pulsating warm moist wind, sexual sound effect

  5. Maybe Foreman is using ‘analogous’ to mean ‘similar’ or is restricting the analogy.

    Perhaps Foreman’s claim is more nuanced than what Hat quotes above. But all that post seems to be saying is phonemic vowel length is disappearing in Korean. We know phonemic vowel length has been lost in Latin and Greek. If that’s the totality of the claim, I see nothing to call ‘analogous’. Particularly, Foreman’s “must have looked like” is the shibboleth for he’s winging it. Is he saying anything more than that in all three languages vowel length used to be phonemic and now it isn’t? (Or rather in Korean it nearly isn’t/it’s vanishing in front of our ears.)

    What do we know about the ‘process’ in antiquity? As recently exposed, my knowledge of Latin vowel length is abysmal. wp has

    The Classical vowel length system faded in later Latin and ceased to be phonemic in Romance, having been replaced by contrasts in vowel quality. Consonant length, however, remains contrastive in much of Italo-Romance, cf. Italian nono “ninth” versus nonno “grandfather”.
    … the automatic stress accent of Classical Latin, which was dependent on vowel length, becomes a phonemic one in Ecclesiastical Latin.

    In Greek it’s durned complicated because “vowels and diphthongs varied by dialect”. But I notice in that article how often ‘lengthening’ appears preceded by ”compensatory’ (and there’s a whole ‘nother article on the topic — this seems to be describing how vowel length became phonemic; we’re interested here in it later ceasing to be so). To cherry-pick:

    In Attic, Ionic, and Doric, vowels were usually lengthened when a following consonant was lost[**].
    … By the 4th century AD poetry was normally written using stress-based metres, suggesting that the distinctions between long and short vowels had been lost by then, and the pitch accent had been replaced by a stress accent.

    So for Korean changes to show some “analogous processes”: replaced by changes in vowel quality like in Latin?; or changes in accent pattern like in Greek? Or both? (Is that a possible thing?) Or something else not analogous?

    ‘Compensatory’ is featuring strongly in my thinking: if a language is going to lose some systematic feature, it needs to gain some other feature so as to maintain a balance of phonemic density per semantic density.

    [**] Losing a following consonant and getting that loss compensated in vowel quality (tone/length) is what I would call ‘analogous’ to what’s inferred happened in Old -> Middle Chinese.

  6. I’m always bemused by these claims some phonemic (sub-)system is “disappearing” or a language is ‘losing’ some distinction. As if there’s a sort of language entropy/all languages will end up using a handful of phonemes.

    I’m bemused by your bemusement. What on earth does the first (something disappearing) have to do with the second (entropy)? It’s as if I said “I lost my wallet” and you said “No you didn’t, because you’ll get another one.” If phonemic vowel length is disappearing, it’s disappearing, and whatever compensatory mechanisms are occurring are irrelevant to that fact. Also, your “if a language is going to lose some systematic feature, it needs to gain some other feature” sounds like the kind of quasi-religious dogma (cf. “all sound change is regular”) that gets people in trouble. It’s one thing to say regularity or compensation is in general applicable, but once you elevate it into a dogma you lose sight of inconvenient facts and it can only end in tears.

  7. No, not dogma: it’s observation — you’ll need to read the tangled story of Greek at that link for example, or the separate article on ‘compensatory lengthening’.

    But it also makes sense in terms of information content: if a language can make n^m distinctions but then loses some phonemic resource such that it can only make n^(m – 5) distinctions, folk will find some other way to restore to n^m approx. Perhaps they’ll introduce a whole load of auxiliary particles rather than use inflections. This is of a piece with combatting those silly claims some language is ‘more expressive’ than another.

    So what I’m saying wrt Korean, is that I’d expect some compensatory effect. The quote from Foreman hasn’t told what. And until we know what compensatory effect, the claim for ‘analogous’ to Latin/Greek is vacuous.

  8. “…folk will find some other way…,”
    or folk may find some other way,
    or do without?

    I really liked the passage on Hindi and Urdu verses.

  9. I follow him on Twitter and judging from those excerpts he posts similar content there and on Facebook.

  10. I was unaware that people post actual blog type content on Facebook. I thought that site was exclusively for vacation photos, photos of your children, promoting one’s business and the occasional political rant. Oh, and, sadly, informing people of a grave illness and perhaps setting up a gofundme. But that may just be my aging GenX demographic.

  11. David Marjanović says

    A political rant that is long enough quickly becomes blog-type content.

    The Latin distinction of /a/ and /aː/ is completely lost, with no compensation, in all Romance languages. To find it preserved, you need to look at words that got out of Latin early enough, mostly into Welsh.

    The English distinction of w and wh is lost without compensation in most accents; in the US, people who have it all seem to be noticeably older than me (e.g. Sen. Elizabeth Warren) and seem to be sprinkled all over the country.

    Sometimes, a phonemic distinction bears little enough information that it’s possible to just live with a few more homophones.

  12. I am a speaker of Standard Korean born in the 1980s and I have no trace of phonemic length in my idiolect. Of course, I learned in school which words were supposed to be pronounced with long vowels and which ones with short vowels, but it was something I had to memorize because I never made those distinctions myself.

    So I have to deal with many more homophones than more conservative speakers who maintain phonemic length, but I live with it, and find ways to avoid ambiguous formulations. I’m fairly sure I haven’t developed a compensatory distinction elsewhere in my phonological system to make up for my lack of phonemic length.

    I use Facebook as my “blogging” platform for the moment (in Korean), by the way. The blogging platform I originally started on closed down this year, but I jumped ship to Facebook several years ago already.

  13. It’s quite easy to lose a phonemic distinction without gaining any new one in the process, and it happens all the time. If there’s any substance to the idea that complexity stays constant (a claim for which there’s precious little evidence), it would have to be over a much longer time scale: not “each loss must be accompanied by a gain”, but rather “over any given millennium, the losses will be roughly counterbalanced on average by the gains”.

  14. I wonder then to what extent it is true for dialects. I understand that if you pronounce unstressed o’s clearly in Russain, then you are not supposed to post in the Internet and must be too busy with your goats (I even saw someone in youtube chat asking “why each time I meet a Russain here, he happens to be from Moscow?”), but Korea is two countries….

  15. Stu Clayton says

    I recently talked with a Muscovite who, on my enquiring, told me that in Moscow “о” as in москва is *always* pronounced as “а”, even in stressed syllables. I would have to go out into the country, he said, to find “о” pronounced as [however the appropriate IPA o-like sound is denoted]. Is that what the goats are about ?

  16. @Stu, goats are about the situation, when certain (widespread) dialects are invisible when you look from Moscow. You don’t hear them on Moscow streets, you don’t hear them on TV (but East Ukrainian Russian was visible or rather audible even before the war) not even on Youtube (where you too hear it). But then you go to a town not so far from here and everyone has clear /o/…
    I don’t know how to explain this.

    Moscow o when stressed is o. DM hears a diphtong there, but it is surely labialised and definitely not open.

    The main (least detailed) partition of Russian dialects is “aking” and “oking” dialects. In the former group a/o (and also usually e/i) merge when unstressed, the result may sound a-like.

    Aking is South-West (extending to Belarusian but not Ukrainian), oking is North-East. The line actually crosses the Moscow region (though here oking only occurs in speech of villagers).

    Muscovites reduce everything unstressed and lenghten stressed vowels. There are two grades of reduction. The second grade is a schwa… Either of the two schwas (more fronted for e/i, more back for a/o). You aren’t going to have palatalised and velarised rows of consonants without having two schwas.
    The first grade – the first syllable when it precedes the stress – found in Moscow for a/o is variously transcribed, Wiktionary uses the upside-down “a”. I would not call it terribly a-like (or maybe it would be so if were less reduced…:-)), there are accents where it is more a-ish.
    But my expectation (I never actually checked) is that we are good at reducing vowels instead.

    ____
    All right, I was wrong about North Korea.

    Ethnologue: “Korean is the official national language of South Korea. It belongs to the Koreanic language family. It is used as a language of instruction in education.”
    twitter.com/theEthnologue: “The most authoritative resource on world languages.”

    Goats then….

  17. Drasvi, Foreman mentioned the stressed /o/ diphthongization recently, see here:

    https://twitter.com/azforeman/status/1665226591071240193

  18. Looks like a partly labialised semi-long vowel…
    (I mean, not much change other than relaxation of lips)

  19. Quite a few linguistics textbooks claim that all languages are equally complex. I assume this is on part a reaction to old biases about since languages being more ‘primitive’ than others, but it’s hardly self evident. I provisionally assume that all languages are *complex enough* to do everything a human language should, but I’d leave room for a bit of real variation in complexity. I’ve sometimes thought of this as all languages having 98-100% complexity, where the upper end of the most complex language ever spoken (with rich phonology, intricate and irregular morphosyntax, large lexicon, etc). Even the simplest language is still quite something. This might not be a useful perspective, but it seems like an improvement on the usual textbook line, at least.

  20. Stu Clayton says

    What purposes are served by these discussions about “complexity” as a property of languages ?

    Is there a hope that, through these discussions, the participants will acquire knowledge that helps them to …. do what exactly ? Produce more or less complex sentences or thoughts ? Or merely to increase one’s personal stock of knowledge ?

    This focus on complexity is pretty simplistic.

  21. Korean dialects can be divided into those that have tones or pitch accents (“tone dialects”), those that have phonemic vowel length (“length dialects”), and those that have neither.

    The earliest texts in the Korean alphabet from the fifteenth century used dots to mark tones, so at least that variety of Middle Korean (presumably from the Central dialect that includes the speech of Seoul) had tones. But the tone markings disappeared very early, and Central Korean is traditionally classified as a length dialect.

    Traditionally, tone dialects were found in the Northeast and the Southeast. Not sure about the Northeastern dialect since it’s part of North Korea today, but tones are still robust in the Southeastern dialect.

    Towards the northern part of the Central dialect region and into many areas of the Northwest, there were many areas that had neither tone nor phonemic vowel length. The dialect (or language) of Jeju Island to the south also has neither.

    But for most speakers of Central Korean as well as virtually all Southwestern dialect speakers, the traditional picture was that phonemic vowel length existed. It naturally became codified as part of Standard Korean pronunciation.

    Today, phonemic length is disappearing from all traditional length dialects, at least in South Korea. Even in the Southwest, I understand that younger speakers are losing the length distinction. I would say that the loss is virtually complete in younger speakers of Seoul (including myself).

    As for North Korea, Pyongyang is traditionally classified as being part of the length dialect area, but there are studies from the 1990s that found that speakers from Pyongyang had virtually abandoned phonemic vowel length, whereas speakers from Seoul were still making some marginal distinctions at that point.

  22. @Jongseong Park, thank you!

    So as i understand, pitch is still there for tone dialects.

    In many languages there is some sort of koine, and there are also numerous “old” dialects that in 20th-21st century experience extreme levelling influence from this koine. When people say “in [language X]” and omit information about dialects, it can be because dialects are invisible for them and are expected to disappear or follow the koine.
    But it is not the situation of Korea (even the South), and especially, North Korea must have different koine, somewhat different prestige variety etc. So i was surprised to read “in Korean” without details. (also for personal reasons, both Koreas are prominently present in my mental map)

  23. It’s just curiosity about language, and how we might think about linguistic diversity and variation. I don’t expect everyone to share that curiosity, but I know at least some other people are interested in that kind of question. If that warrants a ‘merely’, then I guess I can’t argue with that.

    This seems like the wrong blog to raise the lack of practical applications as an objection to talking about something!

  24. This seems like the wrong blog to raise the lack of practical applications as an objection to talking about something!

    You mustn’t mind Stu, he takes every opportunity to point out how meaningless and pointless everything is. Just think of it as a refreshing blast of nihilism wafting through the room from time to time. (He’s now going to tell me I’ve completely misinterpreted his thinking, and if I would only read Luhmann it would all become clear…)

  25. Lars Mathiesen (he/him/his) says

    But everything is meaningless and pointless, that’s what makes it so interesting. Because it’s the human kind of meaningless and pointless, and we’re humans.

  26. @David M Sometimes, a phonemic distinction bears little enough information that it’s possible to just live with a few more homophones.

    Sure. Loss of English ‘whom’ vs ‘who’ — whose rules of ‘correct’ application were always peevery. Foreman isn’t talking about one or two words, but loss of phonemic vowel length across all vowels/all words (presumably). This’ll produce a lot of homophones, I’m expecting; but I’ll await the data (see below).

    @Lameen the idea that complexity stays constant (a claim for which there’s precious little evidence) …

    I presume you’re responding to something I wrote. But I didn’t mention complexity, and made no claim about it. Then …

    @Stu What purposes are served by these discussions about “complexity” as a property of languages ?

    I see no discussion about “complexity”. Or indeed about complexity. Certainly not from me. You’re very welcome to be nihilistic (in @Hat’s attribution); but even a nihilist must identify their target first. (And I’m equally welcome to just ignore nihilism.)

    Today, phonemic length is disappearing from all traditional length dialects, at least in South Korea.

    Thanks @Jongseong, for volunteering some data! So, are there now particles/words/phrases that are homophones where previously they sounded distinct? Does the writing system still make that distinction?

    Do speakers sometimes produce exaggerated (what English calls) ‘spelling pronunciations’ to disambiguate where there’s a risk of confusion?

    Or do speakers use other linguistic means to avoid ambiguity? Or is there an increase in misunderstandings, needing speakers to supplement what they just said after seeing their listeners bewildered?

  27. @drasvi But it is not the situation of Korea (even the South), and especially, North Korea must have different koine, somewhat different prestige variety etc. So i was surprised to read “in Korean” without details.

    Interesting. Would you have the same reaction if someone were to talk about changes happening in English, French, Mandarin Chinese, or Russian without specifying which variety?

    @AntC So, are there now particles/words/phrases that are homophones where previously they sounded distinct? Does the writing system still make that distinction?

    Korean orthography has never marked length. As I mentioned, tone was marked for a brief time when the alphabet was invented, but it stopped being used pretty quickly.

    There are plenty of words that are homophones for me and most other speakers of Central Korean because we no longer distinguish phonemic vowel length:
    gul “oyster”, 굴 guːl “cave”
    nun “eye”, 눈 nuːn “snow”
    mal “horse”, 말 maːl “speech”
    bam “night”, 밤 baːm “chestnut”
    beol “punishment”, 벌 beoːl “bee”
    byeong “bottle”, 병 byeoːng “illness”
    사과 sagwa “apple”, 사과 saːgwa “apology”

    However, in most cases it is possible to tell which meaning is intended from context. In rare cases where we do need to specify, we can say things like 먹는 밤 meongneun bambam that you eat (i.e. chestnut)”. There is no way to pronounce them in an exaggerated manner to differentiate them, because without a length distinction, these forms are identical (as you can tell by the spelling). If someone pronounced byeong with an exaggerated long vowel I would still have no idea what they meant without looking up which meaning of byeong gets the long vowel.

    It’s pretty much the same situation as for all the homophones created by mergers in English such as “Wales” and “whales” for most speakers, or “father” and “farther” for non-rhotic ones, except that in English you see them distinguished in spelling.

    Only about 15% of all words used long vowels in Korean. Long vowels could only occur in the first syllable of a word, so even words that were pronounced with long vowels lost them when they were used as non-initial elements in compound words. Of course, this might be because phonemic vowel length had already been in retreat for a long time (or was the vestige of a tone system that had long disappeared), but the point is that the functional load of phonemic vowel length was not as high as one might first suppose.

  28. @Jongseong Park,

    In Russian – not at all.
    In English – I would be very much surprised: people are generally aware of differneces in spelling and usage between varieties of English, and native speakers keep surprising me with their knowlege (and curiousity about) phonological variation…
    Mandarin is tricky, because neither “Chinese” (a large family) nor “Mandarin Chinese” (usually understood as “the promoted variety as opposed to other varieties” rather than “a group of northern dialects”) are analogous to “Korean”
    French… Strangely I don’t know enough about how people see/perceive “French” (even though I know quite a few native speakers and L2 speakers who use it daily).
    (PS: similarly, I would be surprised to hear it about “Arabic”, but not about “Moroccan Arabic”, even though there are dialects in Morocco. Usually people mean the koine (“darija”) and dialects share many traits)

    “tone was marked for a brief time when the alphabet was invented, but it stopped being used pretty quickly” – I didn’t know that!

  29. Similarly, I was surprised to find that stress used to be marked in Russian.

  30. @drasvi,

    I would have thought Mandarin Chinese (by which I mean the northern dialect group of which Standard Mandarin is a subset) was the closest analogue to Korean out of the examples I listed.

    China and Taiwan both use the same standard language under different names, with minor differences in vocabulary and pronunciation. You also have lots of speakers of dialects of Mandarin that differ from the promoted standard.

    Granted, both China and Taiwan have large numbers of speakers of languages other than Mandarin while the Koreas don’t have any indigenous language other than Korean, and we can’t forget the Mandarin spoken in Singapore, but I would have thought this was a pretty clear parallel to the situation in the Korean peninsula.

  31. David Eddyshaw says

    Plenty of phonetic mergers have happened in the Oti-Volta languages.

    In Kusaal siig “life force, spirit” is now completely homophonous with siigAnogeissus leiocarpa, African birch tree” (in Buli, respectively chiik and siik.)

    Nobody seems to be confused that the Third Person of the Trinity is called the Holy Birch, though.

  32. Athel Cornish-Bowden says

    Similarly, I was surprised to find that stress used to be marked in Russian.

    Still is, is it not, in the Bible and in books for foreigners and chidren?

  33. Athel Cornish-Bowden says

    Quite a few linguistics textbooks claim that all languages are equally complex.

    Is this something that you have to swear to before being allowed to take a linguistics course, rather as trainee priest has to manifest a belief in Christ’s divinity? For someone who hasn’t taken such a course it seems absurd.

    Does anyone seriously claim that Malay is as complex as Navajo? Or that Welsh is as complex as Russian? Of course, learning to speak either Malay or Welsh like a native speaker may take a lifetime, but learning to make oneself understood at a basic level does not. Can one say that of Navajo or Russian?

  34. Still is, is it not, in the Bible and in books for foreigners and chidren?

    Not in the Bible, no. The point is that it used to be marked regularly, in books for ordinary Russian speakers. (We’re talking several centuries ago.)

  35. For someone who hasn’t taken such a course it seems absurd.

    With respect, lots of things seem absurd if you don’t know anything about the topic. Once upon a time, it seemed absurd to say the earth went around the sun. All languages are complex, but the complexity appears in different aspects; as I said here:

    Yes, it’s an exaggeration, but it started for a good reason (because racists kept claiming primitive people spoke primitive languages), the fact that it’s an exaggeration is pretty obvious on the face of it, and to attack it with such force (given that it plays no part whatever in the actual work of linguistics) makes one look as picky as the people who insist on demolishing the statement that “all men are created equal.”

  36. Athel Cornish-Bowden says

    I find it odd that modern Russian (and English) don’t mark stress. It’s very useful in Spanish, though getting the stress wrong in Spanish is not usually a disaster. Something I found surprising when I tried to learn Russian about 60 years ago is that textbooks emphasize the importance of stress and say that it’s not easy to guess which syllable should be stressed, but omit to mention that much the same can be said of English, so from the point of view of an English speaker it’s not a bizarre characteristic of Russian.

  37. David Eddyshaw says

    Or that Welsh is as complex as Russian?

    The Russian initial consonant mutations are a real headache. To say nothing of all those plural forms you just have to memorise individually. And the VSO word order and conjugated prepositions are likely to cause problems for someone who has never been exposed to a Slavonic language before.

  38. @LH, consider
    “all languages are equally adorable”
    “all langauges are equally beautiful”

    I will not interpret these two as scientific bullshit. I will interpret them as an expression of attitude. Why?
    For ‘adorable’, ‘admirable’ and so on we beleive that these things should be subjective.
    As for ‘beautiful’, we think that beauty is outside of the sphere interests of science.

    Is this right? Possibly not. Possibly sceince could try to understand beauty. I would not discourage one from doing it. There are many things that were outside of the sphere of interest of science before, but are science now (including language).

    Compared to this, claims about complexity are firmly within the sphere of interests of science.

  39. I didn’t say they weren’t, and I’m not sure where you’re getting that. I’m saying that while it’s an oversimplification to say “all languages are equally complex,” it’s a reasonable first approximation, useful in countering the layman’s attitude that certain languages are “obviously” more complex than others.

  40. ktschwarz says

    What textbooks should say, I think, is that all languages are so complex that we don’t even know how to describe how complex they are. Can ants determine whether one mountain is higher than another? All languages are so complex that a tome the size of CGEL just barely scratches the surface of what people do in everyday conversation.

  41. @LH, because the only way I can understand

    and to attack it with such force … makes one look as picky as the people who…

    is that you read “all alngauges are equally complex” as an expression of attitude (and can’t imagine how anyone could read it as a (false) scientific claim).
    I, in turn, don’t really understand how this can express an attitude – I love Afrikaans, it is easy – but I read it as obvious scientific nonsense. And textbooks are not a good place for bullshit.

    (having this said, I don’t rememeber ever seeing this claim in a textbook).

  42. No, you’re misreading me.

  43. David Eddyshaw says

    It obviously depends on how you define “complex”, but even so, languages do vary tremendously not only in obvious things like phonological and morphological complexity but syntactic complexity too. You’d have to make up some very idiosyncratic definitions of “complexity” to make them all come out much the same. And I just don’t think it’s true that simplicity in one grammatical domain is balanced by complexity in another. It just isn’t, except in the very limited sense that comparatively isolating languages have to do with syntax what morphologically more exuberant languages do with flexion.

    Languages also vary hugely in morphological unpredictability. I mean, in Kusaal you form the imperfective aspect by adding the same suffix to all the verbs in the language except one. Russian aspect formation is a bit less predictable IIRC …

  44. I think I’ve brought up Latin and (Classical) Greek before as disproving the equicomplexity hypothesis — very similar in structure, but Greek is more complex and irregular in just about every part of the grammar.

  45. David Marjanović says

    Compared to this, claims about complexity are firmly within the sphere of interests of science.

    Ha! All attempts to measure the complexity of an organism have been abandoned – there’s simply no scientific criterion for what to count and how to weight it.

    The point is that it used to be marked regularly, in books for ordinary Russian speakers. (We’re talking several centuries ago.)

    Rather, Cyrillic used to contain the whole apparatus of Greek accents and tried to use it. Of course Greek itself had lost its pitch accent centuries before Cyrillic developed (and further centuries before Cyrillic was perceived as something separate and not just “Greek script with a bunch of extra letters”), so it was all just ornamental, except for the side effect of marking where the stress goes (which is almost always in the same places in Church Slavonic as in Russian, usually one syllable later than in Serbian). Peter the Great had enough of this and simply abolished all the squiggles. Because he was Great, the entire Cyrillic-writing world followed him pretty quickly.

    That’s actually similar to the Korean situation: phonemic length was a remnant of the tone system – it corresponds 1 : 1 to the rising tone of Middle-Korean-as-we-know-it. In the length-only dialects, the high and the low tone merged, but their associated shortness continued to distinguish them from the rising tone, which ceased to be rising…

    In modern Russian text I’ve once or twice seen accents used to distinguish certain forms of “larger” and “more”. But not otherwise, apart from Wikipedia headwords and the first few lessons and the vocabulary lists in my schoolbook.

  46. Oh, it’s used more than that — e.g., to distinguish доро́гой ‘on the way’ from the adjective ‘dear.’

  47. John Cowan says

    Peter the Great had enough of this and simply abolished all the squiggles.

    If he had been truly Great he would have done what the Greeks did later and abolished all but one. Still, the Greeks could have gone further and abolished all but proparoxytone/properispomenon accents.

  48. @LH, obviously I am. The fact is that we read this claim differently, for you it is imprecise but acceptable, for me it is not.

    I know one scenario, when I can accept imprecise claims in textbooks, I described it. I can invent more, but maybe you could help.

    We understand the claim differently and perhaps it is not easier for you to understand why, than it is for me. But at least I don’t think that your idea of scientific honesty “looks” as loose as that of “the people who…” etc. I just don’t undestand you.

  49. Also, can’t we defend views of those guys from 19th century the same way?

    “Analytical languages are simpler and reflect simple thinking”
    “So what about English?”
    “Ah, don’t be picky. You know what I mean”.

  50. Andrej Bjelaković says

    Re the complexity stuff, I’d recommend this podcast with Peter Trudgill:

    https://hiphilangsci.net/2022/08/01/podcast-episode-27/

  51. @LH, another try:
    the problem for me is that instead of “we don’t understand” it is “nothing to understand here”. It discourages study of complexity.

    If you’re afraid that people may think something stupid about complexity, you can warn them without making false claims (I can tell a beard from a woman’s breast without assuming that one gender is “better” or “more dumb” or whatever else other than “more bearded” than other, so maybe we can deal with sexism without abolishing obstetrics. And it is not that the danger is not there, sexism does exist (and it surely wouldn’t if women and men were identical and reproduced by budding)).

  52. the Greeks could have gone further and abolished all but proparoxytone/properispomenon accents

    How would that work — e.g. how would you tell the accent position on a word like ὀλίγος?

    A better way to simplify the system might be to a have a rule that recessive accentuation is default and mark all deviations from that.

  53. John Cowan says

    A better way to simplify the system might be to a have a rule that recessive accentuation is default and mark all deviations from that.

    That’s what I meant (sorry, I confused myself, never mind anyone else). On two-syllable words, put an acute accent on the stressed syllable if it is final. Otherwise, put an acute accent on the stressed syllable unless it is antepenultimate.

  54. You’d still need to keep the circumflex, though, to not lose occasional important distinctions, e.g. μήτις ‘no one’ / μῆτις ‘cunning’, ή ‘or’ / ῆ ‘in truth’ (I’m assuming we’re ditching the smooth breathing too), παιδεύσαι aor. opt. / παιδεῦσαι aor. inf., καλοί (nom. pl. of καλός) / καλοῖ (opt. of καλέω), λαβών (aor. ppl. of λαμβάνω) / λαβῶν (gen. pl. of λαβή).

  55. John Cowan says

    Modern Greek (which is what interests me here) already uses only the acute (called the tonos) and no breathings.

  56. ktschwarz says

    Other useful discussion here about measuring language complexity: Simplification Isn’t Simple (2015), starting with a McWhorter article claiming that languages will get simpler in the future, and Sally Thomason’s response. TR’s comment about Ancient Greek vs. Latin is there.

  57. @Jongseong Park There are plenty of words that are homophones for me and most other speakers of Central Korean because we no longer distinguish phonemic vowel length:

    Thank you. (My q about the writing system was considering Mandarin, where with the collapsing of the number of tones there are many homophones; but speakers ‘think of’ them as distinct lexemes because each is a different character. Except of course it’s a whole lot more nuanced than that.)

    in most cases it is possible to tell which meaning is intended from context. In rare cases where we do need to specify, we can say things like 먹는 밤 meongneun bam “bam that you eat (i.e. chestnut)”.

    Yeah, ok: with those examples you quote, I’m struggling to imagine an utterance that would be ambiguous. Perhaps Korean cunningly organised vocab into distinct topic areas, anticipating loss of lenition. (And English uses similar disambiguation: funny-haha vs funny-peculiar.)

    The situation seems to be much more hazardous in Mandarin: people use speech-to-text to send messages (because no one can remember all those characters); then text-so-speech at the receiving end. Except I have a friend who prefers to use text, who is constantly complaining people are using the wrong character. I also observe many speakers checking the text before sending, by looking up the meanings of the generated characters. There seems a need for frequent re-recording of messages.

    … this might be because phonemic vowel length had already been in retreat for a long time (or was the vestige of a tone system that had long disappeared), but the point is that the functional load of phonemic vowel length was not as high as one might first suppose.

    Thank you again. So diachronically [**] we might need to go back further in time to look for ‘compensatory’ changes in word forms or syntax.

    [**] To those persisting in empty discussion about “complexity”: I was at no point comparing across languages, but only within-language diachronic changes described as ‘compensatory’. I’m making no claim there’s anything to compare Greek wrt Latin wrt Korean. (Do they even have comparable numbers of phonemes? Consonant clusters?)

    ‘Compensatory lengthening’ is enough of a linguistic phenomenon there’s a wiki page for it. ‘Complex*’ does not appear on that page. ‘Compensatory’ means preserving the number of distinct phonemes **within the one language**/presumably to avoid homophones. _Not_ to ensure each language keeps up to some international quota of distinctions.

  58. Keith Ivey says

    funny-haha vs funny-peculiar

    “Writing pen” vs “stick pin” in parts of the US South.

  59. Ah, I always forget about Modern Greek.

  60. “Complex” was an unfortunate choice of words on my part, tying this into a broader debate than AntC had intended. However, the point stands: sound changes normally do not avoid creating homophones. Korean loss of vowel length, but also Songhay loss of tone, Moroccan Arabic short vowel collapse… Such processes simply leave the language with a bunch of inconvenient new homophones to cope with or to deal with by lexical replacement.

  61. Keith Ivey says

    Somehow the western part of the US manages to express collar/caller, Don/Dawn, lager/logger, wok/walk, etc without introducing new phonology, but I suppose the number of pairs that realistically might have contexts they’d be confused in is pretty small.

  62. John Cowan says

    ‘Compensatory’ means preserving the number of distinct phonemes **within the one language**/presumably to avoid homophones.

    It doesn’t even mean that: it means preserving the amount of time it takes to say words. When /x/ was lost in English, /nixt/ ‘night’ became /niːt/ (later /naɪt/ by the Great Vowel Shift); that was compensatory lengthening even though the /i/ vs /iː/ distinction was already present in the language. (In Greek we have the first compeenatory lengthening followed by the second compeesatory lengthening.)

  63. Middle Korean permitted initial consonant clusters of up to three consonants, as in ᄢᅢ다 pskayta “to break, to shatter” (using the Yale romanization here since it’s Middle Korean) corresponding to Modern Korean 깨다 kkaeda.

    Today, Korean syllables are maximally CVC, or underlyingly CVCC where both coda consonants surface only if the following syllable doesn’t have an initial consonant. This means that Korean has clusters of two consonants at most, restricted to between vowels. Some nonstandard pronunciations might have three-consonant clusters. 밟다 (“to step on”), orthographically balbda, is normatively pronounced [b̥aːp.ta] (with a long vowel!) as if written 밥다 bapda, though most people seem to simplify the cluster differently, saying [b̥al.ta] as if written 발따 baltta; intriguingly, there are those who report that they pronounce something like [b̥alp.ta] without the expected reduction.

    Korean has lost some vowel phonemes (such as that represented by the Hangul letter ㆍ which might have been something like [ʌ] or [ɤ]) and have gone though a number of mergers. Vowel mergers are in progress in basically all dialects; there is hardly any speaker who maintains the normative 10-vowel system of Standard Korean. I, like most Central Korean speakers of my generation, have a 7-vowel system in my everyday speech, though I can certainly elicit all the vowels if required.

    All this is to say that Korean has been losing phonological distinctions here and there for centuries. I guess the reason that the resulting homophones don’t pose much of a problem is that in practice, they are unlikely to be ever confused for each other (different parts of speech, etc.) and lexical replacement is likely to take place in the exceptional cases. Also, there have been some developments in the direction of restoring phonological complexity as well. Many more words start with the tense series or aspirated series of consonants in today’s Korean than in Middle Korean for example.

  64. ktschwarz says

    You don’t even need to study any language besides English to observe that lost distinctions often aren’t compensated. Take the NORTH-FORCE merger, aka horse-hoarse: spelled differently because they were (formerly much more widely) pronounced differently. Mary-merry-marry. Vein-vain. Cot-caught and pin-pen, as Keith mentioned. Dozens more, depending on accent. Initial cluster reductions: knot-not. None of them have been compensated by anything, except occasionally lexically when needed, as with “ink pen”. (We do have conservative spelling, but people don’t generally need to resort to spelling to have a conversation.)

    AntC had an unexamined assumption that phonological space is densely filled, that the reduction of n^m distinctions to n^(m – 5) distinctions would cause a catastrophic number of collisions that must be avoided. In fact languages usually don’t make use of anywhere near all of the possible space. I think there have been some calculations of the fraction of phonotactically possible English sequences that are actually words, maybe it’s even been discussed here; whatever the fraction is, it’s significantly less than 1, and decreases sharply as words get longer.

    (Chinese may be different; all I know is from reading Language Log, but my impression is that the space of possible Chinese monosyllables did become too densely filled to work in practice and that’s why most words now have two syllables.)

  65. John Cowan says

    None of them have been compensated by anything

    See loss of /x/ above.

    the space of possible Chinese monosyllables did become too densely filled

    For “Chinese” read “Mandarin”. All the Sinitic languages have lost some of the Middle Chinese distinctions, but Mandarin (which is the French of Sinitic) has gone way over the top. Consequently, 子 ‘child, seed’ is now a suffix meaning ‘(noun)’.

  66. ktschwarz says

    None of them have been compensated in the way that AntC meant “compensated”, with some other distinction appearing in order to avoid the scourge of homophones.

    For “Chinese” read “Mandarin”: OK, thanks.

  67. David Marjanović says

    WP: Standard Chinese phonology § Syllables, links omitted:

    Many of the possible combinations under the above scheme do not actually occur. There are only some 35 final combinations (medial+rime) in actual syllables (see pinyin finals). In all, there are only about 400 different syllables when tone is ignored, and about 1300 when tone is included. This is a far smaller number of distinct syllables than in a language such as English. Since Chinese syllables usually constitute whole words, or at least morphemes, the smallness of the syllable inventory results in large numbers of homophones. However, in Standard Chinese, the average word length is actually almost exactly two syllables, practically eliminating most homophony issues even when tone is disregarded, especially when context is taken into account as well.[15][16] (Still, due to the limited phonetic inventory, homophonic puns in Mandarin Chinese are very common and important in Chinese culture.[17][18])

    For a list of all Standard Chinese syllables (excluding tone and rhotic coda) see the pinyin table or zhuyin table.

    Some of the possible syllables don’t occur because they don’t have a regular source through sound change from Middle Chinese. For example, it’s a great mystery why “pull, draw” is : voiced Middle Chinese initials, like the unchanged l, normally caused the development of the second Mandarin tone instead of the first.

  68. (There is the contrary/reductionist point of view that you don’t need phonemes at all. You can always communicate by pointing/hand gestures/nodding/smiling. The sounds coming out of your mouth mean no more than ‘I want (you to do) something’. Child language acquisition.)

  69. David Eddyshaw says

    The French pronounce ubi and Augustus identically.

    Greetings on the First of Where!

  70. David Marjanović says

    Many have actually restored the t. But, conversely, au mois de is widely used with all months.

  71. For example, it’s a great mystery why “pull, draw” is lā

    Completely ignorant conjecture: Influenced by the command “Pull! Pull! Pull!”

  72. Lars Mathiesen (he/him/his) says

    As previously covered here, that’s how Danish works.

  73. ktschwarz says

    The hiphilangsci interview with Trudgill has been linked here before, and some of its references are required reading on the topic. (Including for AntC, since Trudgill discusses diachronic change in complexity as well as comparisons between languages.) What struck me right off is that “all languages are equally complex” isn’t even on his radar as a myth to be debunked; he takes it for granted that languages vary in complexity and starts right in talking about what social environments favor complexification and what kinds of contact drive languages to become more complex or less complex.

    After that opening, it sounded odd to hear him spout Chomskyan orthodoxy later on: “The structural characteristics of all human languages, I would say, past and present are due to the nature of the common human language faculty.” At least that assumption hasn’t stopped him from recognizing that languages are in fact very different from each other, it just makes him think that that’s something that needs to be explained, that without some “other things” influencing them they’d all be the same.

  74. January First-of-May says

    My q about the writing system was considering Mandarin, where with the collapsing of the number of tones there are many homophones; but speakers ‘think of’ them as distinct lexemes because each is a different character.

    I’ve often had similar thoughts about French [cf. David Eddyshaw’s comment of four days ago] – how do they deal with all the homophones?
    (And AFAIK the dialects mostly tend to have even more mergers.)

  75. @ktschwartz The hiphilangsci interview with Trudgill has been linked here before, and some of its references are required reading on the topic. (Including for AntC, since Trudgill discusses diachronic change in complexity as well as comparisons between languages.)

    I’ve tried, but given up in disgust. Trudgill seems to be making as many evidence-free claims as is Chomsky’s bent.

    Firstly and most annoyingly, there’s no definition of ‘language complexity’ for Trudgill’s purposes. He says polysynthetic languages are notoriously complex. [“There is no generally agreed upon definition of polysynthesis.” wp] Well, maybe from the point of view of anthropologists brought up speaking Indo-European languages. Have we asked somebody brought up speaking Yupik how hard it was to do research on English as she is spoke? CGEL is a monstrous tome; and yet everybody agrees it’s far from complete as a description of one language — with hardly any polysynthesis.

    ‘language contact has been responsible for much reduction in morphology in Europe over the last two millennia’

    Ok. But who’s to say reduction in morphology amounts to ‘less complex’? Why not characterise that change as ‘increased reliance on constituent ordering’? (Including ordering of particles.) And who’s to say which ‘strategy’ is the more ‘complex’?

    Here’s some evidence-free what I’d characterise as made-up bollix:

    languages get more complex unless something else happens,

    If languages are spoken today, and perhaps for the last 2,000 years, are not typical of human languages as they typically have been spoken for most of human history, …

    that small children are brilliant language learners,

    (That last merely aped from Chomsky. Does Trudgill have children? Was he present when they were growing up? ‘Brilliant’ compared to … chimpanzees? … adults? … chatGPT? Just a vacuous claim.)

    We simply don’t know what’s typical for languages spoken ‘for most of human history’. He’s confabulating. Therefore we can’t say they were getting more complex ceteris paribus before the time of writing them down — for some definition of ‘complex’. Floods, famines, climate change, wars, invasions, population movements, genocide were typical for most of human history — we know from the archaeological record. We also know from the past ~4,000 years they have dramatic effects on language change.

    I’ve at least been arguing a far more modest idea: that within one language over time, it remains roughly at a steady level of phonemic resources. One bunch of changes that might increase the risk of collision balances against other changes that reduce that risk. But I’ve made no claims whatever about ‘complexiy’ in general, nor comparing one language to another.

    And a lot of the feedback here has been counter-examples of losing/merging phonemes with nothing balanced against. So, did I miss where Trudgill is saying anything relevant?

  76. @ktschwartz required reading on the topic

    Next time don’t go telling me what’s “required reading”. I suggest you first ascertain what “the topic” is. I kept saying. I repeat. I’m making no observations/claims on ‘complexity’ of language-in-general. It’s vacuous. You’re wasting my time.

  77. @Trudgill features are all developed out of earlier, less complex states. [03:44] So if you take German umlaut, for example, which is a complexity in the Germanic languages, then if you go back far enough into Proto-Germanic, there was no umlaut.

    This is absurdly simplistic to compare with-umlaut vs without-umlaut. Is Trudgill a total idiot?

    Umlaut is a form of assimilation, the process of one speech sound becoming more similar to a nearby sound. Umlaut occurred in order to make words easier to pronounce. [wp]

    “easier to pronounce” = _less_ complex (by one possible measure)

    Why the F is Trudgill alleging Umlautisation is more complex? Oh of course: because he’s not defined ‘complex’; he’s just making it up as he goes along.

    What other changes happened in Germanic languages over that timescale? Were some of them ‘balancing’ against what was happening with Umlaut? Shall I describe them as complexification? or as simplification? I can follow Trudgill’s lead and just draw those judgments out of thin air — to ‘support’ my claim the overall phonemic resources remained about steady.

    I have a feeling I might have read a book chapter/paper or two of Trudgill’s. I might even as an undergraduate have been at some of his Sociology lectures at U of Essex. He didn’t strike me as a complete numpty. (But then neither was he commenting on Linguistics.)

    Complexity note for a non-polysynthetic language: move that subphrase “as an undergraduate” about in the sentence. (Which is what I was doing as I drafted it.) What difference does the position make to the sense and emphasis and cadence? Is English more or less complex than languages with a more rigid word order?

  78. This seems relevant:
    https://scholarspace.manoa.hawaii.edu/server/api/core/bitstreams/39e3c365-2c87-47ef-9392-57fe7bc0f013/content

    One paragraph therefrom:

    To answer this question we need first to discuss what complexity might be. My brief
    answer (see further Trudgill 2011) is – and I acknowledge that there are a number of other
    possible answers – that complexification consists of factors such as: increase in irregularity,
    increase in morphological opacity, increase in syntagmatic redundancy, and increase in
    morphological categories. These are all factors which make for L2 [second language] difficulty
    – linguistic phenomena which are difficult for post-critical-period adult learners to
    acquire.

  79. David Eddyshaw says

    Why the F is Trudgill alleging Umlautisation is more complex?

    The WP definition is basically wrong.

    “Umlaut” (properly so called) is more complicated than what gave rise to it, because sound changes have obliterated its original cause, leaving it opaque.

    German umlaut arose by (not currently “is”) a process where basically back vowels were realised as more front before a /i/ or /j/ in the next syllable. At that stage, you’re just talking about allophones, not actual umlaut.

    Later, the /j/ disappeared, and /i/ fell together in unstressed syllables with all other short vowels, to the original reason for the pronunciation difference fell away, and you’ve got an actual vowel change in the plural but no actual contemporary reason for it any more.

    Parallel in Kusaal: the original forms of the words for “child” and “owl” were

    biiga plural biisɛ
    viugɔ plural viidɛ

    What’s going on here is that a long unrounded vowel gets its second half rounded before a velar consonant followed by a rounded vowel. This is entirely automatic, and there was no other source of diphthongs of this kind. So, actually, you don’t need diphthongs among your vowel phonemes at all: you could just represent these words as

    /bi:ga/ plural /bi:sɛ/
    /vi:gɔ/ plural /vi:dɛ/

    However, Kusaal has in fact lost final short vowels in most contexts, including citation forms, so the actual current Kusaal words are

    biig, plural biis “child”
    viug, plural viid “owl”

    So now you do need diphthongs among your vowel phonemes, so the vowel system has got substantially more complex; and so has noun flexion, because you now need an actual rule to explain the sound change in the singular of “owl”, whereas before, the whole thing was just automatic.

    within one language over time, it remains roughly at a steady level of phonemic resources

    There are any number of counterexamples. Lameen, above, mentioned, for example that Western Songhay languages have lost tone. “Woman” and “ten” used to be pronounced differently; now they’re pronounced the same. The languages just lost tones entirely as a phonemic category, with no kind of compensation.

    Western Oti-Volta lost palatal stops. It just has two fewer phonemes: it hasn’t concocted any replacements. “Life force” and “birch tree” used to be pronounced differently; now they’re not.

    Mooré has lost /k͡p g͡b/ completely, just substituting /k g/ everywhere for them. Two less phonemes …

  80. Thanks @FJ. Trudgill hints at that measure in the transcript. (He mentions Ancient Greek definite article having 36 forms.) And his evidence that pre-literate languages exhibit those varieties of features? … impossible to know — since he’s talking tens or hundreds of thousands of years B.P.. Constituent ordering patterning is not so complex because … why? In short: if English/Indo-European wasn’t his prime mindset, would he emphasise morphology so much?

    increase in irregularity

    Then does Germanic Umlaut increase or decrease regularity? How does he know what a human brain treats as (ir)regular? Does Sinitic languages switching syllable codas from consonant (clusters) to tones make it more or less regular? Oh of course that’s happened within the past 2,000 years, so doesn’t count. He’s proving just about as adept as Chomsky at insulating his blatherings from evidence.

    difficult for post-critical-period adult learners to acquire

    That’s an entirely different (seemingly) empirical claim to be making. So L1 speakers of Ancient Greek would find it much easier to learn English as adults than vice versa? Again in fact untestable. (With all due respect to L1 Russian speakers or even German — if we can take them as examples of a more morphologically complex languages — I see no evidence they find it simple(r) to learn English. What trips them up is word order and the number of ‘filler’ words/particles that might be optional in their language but aren’t in English. I’m not sure I even see L1 English speakers finding it easy to learn English.)

    I’m entirely unconvinced pre-historical human societies were somehow tighter-knit and isolated in a way that promoted language complexity (in Trudgill’s sense). Of course we don’t have linguistic evidence, but Stonehenge and Gobekli Tepe and trade of cultural valuables through the Tarim Basin show long-distance co-ordination and communication.

    Now it’s not that I’m proposing some other measure of ‘complexity’. I repeat: ‘complexity’ as a measure across different languages is vacuous. I was making no claim about it.

  81. You’re wasting my time.

    Could you please try to train yourself out of responding to people like that? You’re acting as if you’re some universally acknowledged genius being pestered by know-nothings. In point of fact, you’re just another bozo on this bus, as likely to be wrong as any of the rest of us. Bluster does not help your case.

  82. David Eddyshaw says

    I think what Trudgill is driving at with “complexity” is primarily opacity. Regular sound changes make what was previously obvious and in need of no special memory work for either L1 or L2 learners unobvious and exceptional.

    Kusaal again:

    Li anɛ mɛligim. “It’s dew.”
    Li ka’ mɛligim. “it’s not dew.”

    Li anɛ gbigim. “It’s a lion.”
    Li ka’ gbigimnɛ. “It’s not a lion.”

    In Kusaal, you just have to know which nouns ending in m do which. In Mampruli, there’s nothing to remember:

    Di nyɛ la mɛ’alim.
    Di pa mɛ’alim.

    Di nyɛ la gbigimni.
    Di pa gbigimni.

    Some day, the Kusaasi may decide that all nouns add when final in a clause with a negation, and simplicity will return to this subsystem. (Actually, this is beginning to happen, a little: the final-negative forms are still largely “historically correct”, but a few have already been reworked by analogy.)

    Trudgill’s point is that this particular kind of complexification is very common cross-linguistically, where a regular sound change obscures a previously transparent pattern. Order may eventually be restored by analogy and levelling. Or not.

  83. @DE “Umlaut” (properly so called) is more complicated than what gave rise to it, because sound changes have obliterated its original cause, leaving it opaque.

    Thanks David, your post arrived as I was composing mine questioning Umlaut’s effect on regularity. (And it’s now late here, I’m not going to try to respond in detail.)

    Do speakers learning a language today trouble themselves with “original cause”s? Do they when speaking mentally go through ‘oh, there’s no longer a final short vowel to drive this, I’d better figure out a diphthong’? You seem to be talking about regularity in the linguist’s rules describing the language.

    If I’m following your description of Kusaal as-was, there’s an ordering in the way the rules apply(?) The speaker needs to flexion the trailing vowel to get the plural form; then spot there’s a velar consonant followed by a rounded vowel; then work backwards to apply the diphthong. Doesn’t sound any simpler than remembering the diphthong in the first place and dropping the trailing vowel.

    Speaking as a programmer, it’s much easier (dare I say ‘simpler’?) to have a long lookup table for singular <-> plural, rather than try to represent some sequence-dependent morphing rule. Memory is cheap; algorithms cause bugs.

    no special memory work you say in the later message.

    So there’s a metric for what adds to ‘memory work’? And evidence speakers prefer less of it?

  84. Stu Clayton says

    Bluster does not help your case.

    Not on this blog. Chomsky, Trump et al get a lot of mileage out of bluster at other venues. Depends on the audience.

    They say the meek and thoughtful will inherit the earth. Well, they better have good lawyers is all I can say.

  85. David Eddyshaw says

    So there’s a metric for what adds to ‘memory work’? And evidence speakers prefer less of it?

    Sure: “something to remember” involves more memory work than “nothing to remember.” Elaborate metrics seem … superfluous in this context.

    Doesn’t sound any simpler than remembering the diphthong in the first place and dropping the trailing vowel

    I didn’t explain it properly. At the stage where the final vowels were still present, there were no diphthongs, phonemically. The diphthongisation was subphonemic. Still, I suppose you could argue that distributing allophones correctly is linguistic “knowledge” possessed by L1 speakers, albeit unconscious knowledge. (And painfully to be acquired consciously by L2 speakers if they want to learn to talk proper.)

    Still, the process could (at that stage) be covered by one single exceptionless phonological rule applied (consciously or not) to all words in the language. After the loss of final short vowels, you need a whole lot more rules, affected by the morphology of the words in question: not only more rules, but more complicated kinds of rules, where you have to specify exactly what sort of word they apply to. That’s surely more complex in absolute terms, metrics be damned.

  86. Stu Clayton says

    Memory is cheap; algorithms cause bugs.

    Computer memory relies on hardware behavior, error-correcting codes, backup, mirroring, power lines etc. RDBMS are very non-trivial contraptions, for example. Memory is an amalgam of algorithms and atoms, use at your own risk.

    Anyway, to contrast “cheap” with “buggy” makes little sense. Flies are cheap and by definition buggy.

  87. @Hat You’re acting as if you’re some universally acknowledged genius being pestered by know-nothings.

    I’m (re)acting as if I’ve just been told … are required reading on the topic. (Including for AntC … .

    I rather think I’m the one who’s making the I know-nothing claims. I’m desperately trying to avoid ‘knowing’ anything about relative ‘complexity’. It’s all together too ineffable. (I was making claims about morphemes.)

    … bluster … Chomsky

    I know-nothing about languages of tens or hundreds of thousands of years ago. Apparently Trudgill claims to.

  88. PlasticPaddy says

    @de
    Re allophones, I think you could say there is a pre-stage where they are in free variation. The first increase in complexity comes when they are frozen to specific environments. The second increase comes when the environmental trigger is removed or reshaped in a way the allophonic phoneme is now “unpredictable”. To give a hypothetical example in English, suppose sp and sb are in free variation. Then suppose only speak/spit but sbout/Sbarta are correct. Then suppose initial s is lost in all these words. You now have ‘peak but ‘boke and ‘boken, which then is a new strong verb paradigm…

  89. David Eddyshaw says

    Surely the whole shtick of allophones is that they do occur in predictable environments, otherwise they’re not actually allophones: whereas free variation is just that: free.

    Still, I suppose you’re right: allophones may develop by parcellng out of previously free variants between different environments, so they’re not free any more,

    Of course, you have the additional complication that an allophone of one vowel in one environment may be phonetically identical to an allophone of some other vowel in a different environment. Nor is it always a trivial matter to determine whether different sounds which never occur in the same environment (e,g. English [h] and [ŋ]) are “allophones” (or even if the question means anything.)

  90. David Eddyshaw says

    @PP: my reply to you has been eaten by Akismet (probably because I mistyped my email address on my phone), but it is a small loss. I was vaguely agreeing with you, and muddying the waters by saying that “allophone” is quite a slippery concept in practice.

  91. Stu Clayton says

    Speaking as a programmer, it’s much easier (dare I say ‘simpler’?) to have a long lookup table for singular plural, rather than try to represent some sequence-dependent morphing rule. Memory is cheap; algorithms cause bugs.

    I agree with the first sentence, and will make bold to call singular-plural table lookup simpler. However, it’s not so much “memory” that makes it simpler, nor even the “lookup algorithm”, simple as it is. Libraries already have those.

    Instead, it’s computer speed that makes the whole business preferable to a lookup in all the rule books in the Library of Babel. You can even add a table index to get a slightly less simple algorithm that makes searches even faster – like the card catalogue in the Library of Babel does.

    The kinds of linguistic “rule” DE seems to be talking about are algorithms of the pre-computer era. They are attempts to provide answers to questions by avoiding straight searches, which are time-consuming.

  92. Stu Clayton says

    The material substrate of “memory” has been expanded in ways that make the old techniques of handling it no longer efficient. That was all that mattered anyway, except for those who made their living by mastering those techniques, efficient or not.

    Programmers know that search algorithms for data on fixed disks (locate sector and track) are very inefficient for data in memory, where there are no sectors or tracks. But books have sectors and tracks, which is why they are superior to scrolls when searching for something.

  93. post-critical-period

    i continue to suspect that the whole notion of a separate “stage” of ability to learn languages is mostly an artifact of infants and very young children experiencing a much larger quantity of exposure to speech (with fewer other obligations or activities) than later learners, naturalized & made scientific-sounding these days through the chomskian fantasy of a “language organ” (four manuals, one pedalboard, as many stops as you need for the epicycles, but only one pipe).

    which makes me skeptical about using “difficulty of learning” as a proxy for “complexity.

  94. David Eddyshaw says

    One of the many questionable “facts” asserted as obviously true by the Chomskyites is the idea that all L1 learners not afflicted by pathological conditions all end up learning their L1 “perfectly.”

    In fact, there is plenty of evidence in English that adults vary quite a bit in their command of more complex syntax, and in polysynthetic languages it seems to be not uncommon for complex polysynthesis to be highly valued aesthetically, the domain of poets and other particular experts in language deployment.

    In fact, all you can really say is that most people not afflicted by a specific disease affecting cognition learn their L1 adequately for whatever uses they have of it.

    So I don’t think, even on the assumption that L1 and L2 learning are essentially the same, and even using learnability as your metric, the fact that most children do in fact learn to speak well enough can be used to justify the idea that languages don’t vary in complexity.

    Though it may well tell us that the significant complexity is not actually in the places we thought it was: Forty cases? No problem! All verbs irregular? Big deal!

  95. David Marjanović says

    “Umlaut” (properly so called) is more complicated than what gave rise to it, because sound changes have obliterated its original cause, leaving it opaque.

    It’s even worse.

    1) It is pretty common for sound changes to become morphologized immediately. Apparently that can happen before their products are even phonologized ( = producing new phonemes or even just changing the phonotactics). Umlaut triggered by *i provides two examples from German alone. One is all the *i-stem nouns: the vowel in the second syllable used to be *i, so you expect to find an umlaut product in the whole paradigm, and that’s exactly what happened in English (guest from *ɣasti- is an example) and Norse. But in German, the oldest written occurrences had already eliminated the umlaut and its triggering *i from the entire singular, using it as a plural marker by analogy to nouns where an *i or *j only appeared in the plural. The second example is all the *es-stem nouns in PIE terms: their suffix, roughly *-os- in the singular and *-es- in the plural, turned into Proto-Germanic *-az- and *-iz-, giving a grand total of nine OHG nouns a plural in umlaut-triggering -ir. And then, the umlaut with its trigger began to spread wildly. It still hasn’t quite stopped. It has spread to all the neuters that were identical to the singular in OHG and still are in the North Germanic languages, it has spread to the *i-stem (Mann, pl. Männer), to the abovementioned “book” (Buch, pl. Bücher; a neuter *i-stem, AFAIK)…

    2) Such morphologization can create new phonemes instantly. The classic example is on p. 217 here.

    3) Sound changes, umlaut included, can also create new phonemes instantly when the new sound they create is already in the system. One nice example is the origins of Norse y: both from *u…i/j as in West Germanic, but also from *i…u/w. Once you had a word with y…ju or y…wi, you could no longer tell what exactly the trigger was, and therefore what the y had been. Suddenly [y] was a new phoneme. A similar one is an umlaut phenomenon in German: absent various consonantal “blocking contexts”, *a…i/j turned into [e] right at the beginning of OHG. It remained distinct from /ɛ/, but was it an allophone of /a/? No, never. [e] also had two other origins: *i…a, and *i followed directly by *r (but not by Proto-Germanic *z which had long merged into r by OHG times). There were lots of words with, for example, *-ja- in the second syllable, so you couldn’t tell if the [e] in the first syllable had been *a (umlauted by the *j) or *i (umlauted by the *ja). Instant three-way distinction between /ɛ/, /e/ and /ɪ/, and that in a system that also contained /ɛː/ and /iː/ (plus [æ] and [æː] which were probably still allophones of /a/ and /aː/ in OHG, but not in MHG).

    Do speakers learning a language today trouble themselves with “original cause”s? Do they when speaking mentally go through ‘oh, there’s no longer a final short vowel to drive this, I’d better figure out a diphthong’?

    No, they just have a larger system to learn. A larger system is more complex by definition.

    Forty cases? No problem!

    Yeah. Given the choice between a Finnic language with nigh on thirty cases plus consonant gradation but nice agglutination apart from that, and a Nilotic language with three cases and total chaos, I’d probably find the former much easier. Even apart from the latter’s tones.

  96. which makes me skeptical about using “difficulty of learning” as a proxy for “complexity.

    Thank you @rozele, well put. Seeing as Trudgill neither gives a tight definition for ‘complexity’ [**] nor a metric for ‘difficulty of learning’ [***], there’s still a lot of unanswered questions before he can even start to make a claim.

    [**] Trudgill and other comments here persist in equating ‘complexity’ with morphology, but not mention constituent ordering. Well your call; but don’t imagine you can conclude anything about language-in-general.

    [***] I’m fascinated observing kids growing up in bi- or tri- lingual environments. It does take longer for them to figure out what’s going on; but then suddenly they can distinguish language A from other stuff, and it’s like they re-cast their aural memory and leap forward. A little later they distinguish language B from C; and not much after that they’re making cross-linguistic puns, and sniggering at me (a speaker of language C) for my accent in A. (Fair cop!)

  97. @DM they just have a larger system to learn. A larger system is more complex by definition.

    By what definition given where? This sounds like more of Trudgill’s hand-waving.

    ‘Larger’ meaning ‘more rules’? So we can observe rule-forming inside people’s heads? Notoriously (speaking again as a programmer) you can express the same algorithm as a long sequence of short rules, with an implied drop-through from one non-applying rule to the next (RPG Programming language is good for that); or as a short sequence of long rules — that is with dense pre-conditions (themselves calculating interim results) and manipulations (‘Convoluted C’ famously good = bad for that). I’ve heard RPG programmers claim RPG is simple and “self-documenting”. I’ve heard C programmers claim C is simple, and higher-level languages just add complexity for no benefit. [Cite Martin Richards, originator of BCPL which preceded C.]

    Thanks @Stu for the Programmers’ background. (Not that we’ve any evidence brains work like CPUs.) I’m imagining strictly in-memory table look-ups; not disk seeks. So a larger chunk of memory to look up adds no complexity “by definition”. Whereas more cascading rules (some with exceptions) does.

    Whether endogenous language change gives more or less ‘complexity’ (for whatever definition you’re choosing); we’re still missing any explanation for _why_ a language should change: it’s not giving any more expressivity ex hypothesi; it’s just saying the same old stuff with different sounds. Bell-bottoms are out this decade; chinos are in. They still just keep the wind off your legs.

    @DE there is plenty of evidence in English that adults vary quite a bit in their command of more complex syntax,

    Indeed. And it’s notable that English average sentence length has been getting shorter over a couple of centuries or so. Is that actually language change or merely the proletarianisation of the media bringing to light that most speakers always have had trouble with complex syntax?

    And does the same thought expressed in a series of short sentences with pronominal cross-references count as more or less ‘complex’ than expressing it in one long relative-claused sentence? (The RPG model cp Convoluted C, again.)

  98. @DE “something to remember” involves more memory work than “nothing to remember.” Elaborate metrics seem … superfluous in this context.

    Ah, I see how the trick is worked: a look-up of irregulars/exceptions counts as ‘memory work’. A series of cascading rules counts as ‘nothing to remember’. Rules (apparently) and all their conditions for applying, interactions/precedence of operation don’t need to be remembered(?) [**] Somehow they get Free Parking.

    I’m impressed how much you’ve divined of the inside of peoples’ heads. From … what, exactly?

    Again to compare computer machinery: program code takes up exactly the same variety of memory as ‘data’, including look-up tables. It was Turing’s insight that code is just data that gave us the jump from Babbage’s and Hollerith’s calculating engines to modern computers.

    I’m not aware of any brain research that can distinguish the “something to remember” variety of memory from the “nothing to remember” variety of remembering.

    So this talk of ‘complexity’ seems to refer to the rules drawn up by linguistic analysis. Does the analysis for morphology take up more or less pages of field notes than for constituent ordering? Or do linguists skip the ordering/syntax niceties because they’re too hard to figure out?

    [**] If you were a Chomskyan, I could see how this ‘Free Parking’ would work: the set of possible rules is innate/whatever memory they occupy is already pre-loaded. X-bar or Merge presumably — or was that last decade’s orthodoxy? All the language learner need do is block out the rules that don’t apply in this language.

  99. David Marjanović says

    the *i-stem (Mann, pl. Männer)

    Overedited and still wrong. The “man” word started as an *u-stem, but the Pre-Germanic changes *nw > *nn and, if I’m not even more confused, *uj > *ij made it irregular in Proto-Germanic already.

    By what definition given where?

    One of these, one would hope.

    I’m imagining strictly in-memory table look-ups; not disk seeks. So a larger chunk of memory to look up adds no complexity “by definition”.

    No, by imagination, not by definition. Do you really think human memory is strictly in-memory table look-up? Isn’t it yet another rule to know when the rule applies and when an exception must be looked up instead?

    we’re still missing any explanation for _why_ a language should change: it’s not giving any more expressivity ex hypothesi; it’s just saying the same old stuff with different sounds. Bell-bottoms are out this decade; chinos are in. They still just keep the wind off your legs.

    Some of language change really is fashion. The continuing spread of [ʀ] between the Pyrenees and Russia and across the Atlantic is an obvious example.

    In biology, selection doesn’t cause mutations either. The mutations have their own causes; some of them are even caused directly by Heisenberg uncertainty. Once a mutation is there, then selection can work on it.

    Rules (apparently) and all their conditions for applying, interactions/precedence of operation don’t need to be remembered(?) [**] Somehow they get Free Parking.

    In the first sentence you have a question mark, in the second you assume it’s not there and simply assert what the first sentence would be like without the question mark.

    This is how you end up getting very upset at strawmen once every month or so.

    One rule is one thing to remember; one exception is one thing to remember; one rule with three exceptions is four things to remember. Sound fair?

  100. @DM One of these, one would hope.

    Yeah I think we really don’t want to go there; or the new riddle of ease-of-learning. (Since learnability seems to be one of Trudgill’s posited metrics.)

    Here’s an oblique angle wrt scare-quote ‘complexity’ of description (note the “simpler” and the “more intuitive”):

    Polar coordinates are most appropriate in any context where the phenomenon being considered is inherently tied to direction and length from a center point in a plane, such as spirals. Planar physical systems with bodies moving around a central point, or phenomena originating from a central point, are often simpler and more intuitive to model using polar coordinates. [wp]

    Some disciplines prefer using Polar coordinates; others prefer Cartesian. Because it makes their formulas ‘simpler’. You can of course convert any point or vector between the two systems — YMMV as to how ‘simple’ the conversion is.

    But is either system more ‘simple’ intrinsically? Does a formula usually expressed in Polar but re-expressed in Cartesian explain anything more or less well because it’s got more symbols and operations? Hence my “talk of ‘complexity’ seems to refer to the rules drawn up by linguistic analysis”.

    one rule with three exceptions is four things to remember. Sound fair?

    No because I can choose a different co-ordinate system (grue-bleen vs blue-green) such that that’s one rule with no exceptions, and conversely your claimedly simpler case will then be a complex rule with three exceptions.

    You need also a metric of why one formulation of a rule+exceptions is ‘simpler’/fewer things to remember than some different formulation. You as an observer/not L1 speaker might formulate/store in memory using a different mechanism. (Perfectly competent native speakers have notoriously unreliable beliefs about how their own language works, when asked to give ‘rules’.)

    Do you really think human memory is strictly in-memory table look-up?

    I have no idea. Neither does anybody else AFAICT. Do brains have ‘secondary storage’? How does the speaker get from the information content to the words; how does the listener get from the words to the information content?

    (I was/continue to be ‘upset’ not at strawmen but at being taken to be claiming something about cross-linguistic complexity, despite my continued protestations it’s a pointless pursuit.)

  101. John Cowan says

    “The structural characteristics of all human languages, I would say, past and present are due to the nature of the common human language faculty.”

    That’s not Chomskyan an und für sich. If there were no such common faculty, some of us would have language and others would not, just as some of us eat with chopsticks and others do not, ergo there is no common human chopstick (I started to write “chompstick”) facility.

    Trudgill’s point is that this particular kind of complexification is very common cross-linguistically, where a regular sound change obscures a previously transparent pattern. Order may eventually be restored by analogy and levelling. Or not.

    Sturtevant’s Paradox (1947): sound change is a regular process that creates irregularity, whereas analogy is an irregular process that creates reguality. A nice example is was/were alternation, a product of Verner’s Law, which in German has been leveled to war/waren, and likewise with E lose/lorn, leveled in E to lose/lost, vs. G verlieren/verloren. Not to be confused with Sturtevant’s Law, which is that PIE voiceless stops get geminated in Hittite (S. worked out the details of the Indo-Hittite hypothesis).

    Well, they better have good lawyers is all I can say.

    Lawyers are primarily good listeners rather than good speakers. An experienced lawyer can listen to a client describing his case while paging through a 100-page document and wind up remembering the legally relevant parts of both.

    I agree with the first sentence, and will make bold to call singular-plural table lookup simpler. However, it’s not so much “memory” that makes it simpler, nor even the “lookup algorithm”, simple as it is.

    Of course, it’s only linear search (“check every entry until you find the right one”) that’s simple: faster algorithms are not so simple. In any case, the key fact is that search would handle languages with 50% or 100% suppletive noun plurals just fine, but no such languages exist. Having to remember which affix to apply is apparently much easier than having to know, across all nouns, which root to use.

    Programmers know that search algorithms for data on fixed disks (locate sector and track) are very inefficient for data in memory, where there are no sectors or tracks.

    Eh? B-trees are equally good both in memory and on disk.

    In fact, there is plenty of evidence in English that adults vary quite a bit in their command of more complex syntax

    But is that complexity as such, or is it a matter of whether you can talk like a book or not? I can and generally do, but I’ve probably spent more time reading books than listening to people. Books of course permit greater complexity because you can reread the self-same utterance until you get it, whereas I am constantly plagued by the sort of thing:

    Interlocutor (as heard by me): “urm urm urm urm urm so I want to buy some of that.”

    Bones (me): “What? Why do you want to, again?”

    Interlocutor: “I want to buy some of that, I said.”

    Bones: “Some of what?”

    Interlocutor: That!

    Bones (giving up): Sure, okay. (Clatter, clatter.)

    In biology, selection doesn’t cause mutations either.

    It might in principle. A moderate degree of radiation tolerance could allow an organism to settle a naturally or artificially radioactive area that other organisms could not, in which case the organism would be (pre)selected for mutation.

  102. David Marjanović says

    Sturtevant’s Law, which is that PIE voiceless stops get geminated in Hittite

    Also *s and *h₂. *h₁ didn’t leave written evidence, so there’s no way to tell if the law applied to all voiceless obstruents. In any case, though, it didn’t apply when another consonant followed.

    in which case the organism would be (pre)selected for mutation.

    It would be selected for being able to deal, one way or another, with an increased mutation rate; that’s not selection for a higher mutation rate. But I should have written “selection doesn’t cause any particular mutations”; that’s what I meant. It’s the great contrast to Lamarckism.

    (Radiosynthetic fungi – they use melanin and radioactivity the way plants use chlorophyll and light – are in fact known.)

  103. @JC just as some of us eat with chopsticks and others do not, ergo there is no common human chopstick (…) facility. [faculty?]

    I’m pretty sure there’s a chopstick-faculty learning window that closes early teens, same as language learning.

    I didn’t use chopsticks until late teens. Although I’ve travelled frequently to chopstick-using parts of the world and am adept at rice and noodles, I just can’t for the life of me shell a prawn using only chopsticks. Let alone dig the delicious meat out of a lobster claw. I end up having to use the whole batterie de cuisine, and smeared in sauce.

  104. John Cowan says

    Faculty, yes. I’m not so sure about the window, though I have seen people eat soup with chopsticks (that is, the solid parts).

  105. @David Marjanović: In fact, almost all mutations are probably products of quantum uncertainty effects—meaning that they involve tunneling through classical forbidden configurations. This is actually ubiquitous in many kinds of chemical reactions, but it is especially relevant in a case like DNA replication, in which the catalytic action of a polymerase brings down the activation energy.* While the barrier for the insertion of the correct nucleotide may be brought down to the point at which the process can occur thermally, mutant nucleotides still face an activation barrier that they generally need to tunnel through.

    * Nucleic acid** polymerases also bring down the free energy of activation, of course. They are good at grabbing nucleotides out of solution, which helps overcome the entropic impediment to the reaction proceeding. However, the entropy part is not relevant to the quantum tunneling aspect.

    ** Why isn’t “nucleic acid” just abbreviated “NA”?

  106. Maltese, like other languages in the world, is neither a difficult nor is it an easy language.” (from a Maltese textbook).

    Philosophical.

  107. Lars Mathiesen (he/him/his) says

    The real question is why deoxy-ribo-nucleic-acid isn’t abbreviated dRNA vel sim.

Speak Your Mind

*