We discussed the new Grambank database a few months ago; now the Max Planck Institute for Evolutionary Anthropology reports at Phys.org on a study they’ve done using it:
Languages around the world differ greatly in how many grammatical distinctions they make. This variation is observable even between closely related languages. […] This grammatical distinction in the case system, along with many others, sets Icelandic apart from its closely related sister languages. “One prominent hypothesis about why some languages show more complex grammar than others links grammatical complexity to the social environments in which these languages are used,” says first author Olena Shcherbakova from the Max Planck Institute for Evolutionary Anthropology.
For example, Icelandic is primarily learned and used by the local population of over 350,000 people. Such relatively small isolated communities are also called “societies of intimates.” In contrast, the other Scandinavian countries, located in close proximity to their neighbors, have larger populations with substantial proportions of non-native speakers.
Such communities are known as “societies of strangers.” Many linguists have claimed that languages with more non-native speakers tend to simplify their grammars, as unlike children, adult learners struggle to acquire complex grammatical rules to master the intricacies of their new language.
But is this Icelandic example representative of the striking linguistic diversity worldwide? Researchers at the Max Planck Institute of Evolutionary Anthropology wanted to find out if the grammars of languages tend to evolve simpler when spoken by larger societies of strangers with many non-native speakers.
For their study published in Science Advances, they measured the grammatical complexity of 1,314 languages using data from Grambank—a newly released global database of grammatical features. These complexity scores were compared to variables detailing the number of non-native speakers in these languages.
Language complexity is a hotly debated topic in linguistics, with many different opposing views. “Many of the disagreements are down to differences in how ‘complexity’ is defined,” says Hedvig Skirgård from the Max Planck Institute of Evolutionary Anthropology. “In this study, we improved the methodology by teasing out two distinct measures: fusion (how many affixes verbs and nouns have) and informativity (how many distinctions are made).”
The results show that societies of strangers do not speak less complex languages. “Instead, our study reveals that the variation in grammatical complexity generally accumulates too slowly to adapt to the immediate environment,” states Shcherbakova.
The well-known counterexample to the claim about social environment shaping grammatical complexity is German. German is learned and spoken by a large number of non-native speakers, and yet, it has retained its case system and many other grammatical distinctions.
The study tests the influence of social environment on grammatical complexity, while accounting for the expected similarities arising from both genealogical inheritance and contact. “Our study highlights the significance of using large-scale data and accounting for the influence of inheritance and contact when addressing long-standing questions about the evolution of languages. It shows how received linguistic wisdom can be rigorously tested with the global datasets that are increasingly becoming available,” concludes Simon Greenhill from the University of Auckland.
I don’t expect there will be much agreement on what if anything they’ve actually shown, but it’s an interesting topic and a potentially useful database. Thanks, Bonnie!
Good to see the Max Planck Institute using the awesome powers of the WALS database for good …
By which I obviously mean that, on this occasion, I find their conclusions intuitively plausible.
Whether they’ve actually really demonstrated their conclusions is a whole other issue, though.
Obviously the immediate problem is to come up with a metric for “complexity” that means much. To be fair, there probably is no really satisfactory way of doing this, and they do the best that one can do in the circumstances, by being careful to define what they mean: but their definition seems unlikely to be able to do what they hope. It’s based on how many categories are marked in noun and verb flexion and the degree to which they are fused as opposed to agglutinated; that’s all well and good, but omits the vital dimension of unpredictability.
Moba verbs, for example, inflect only for irrealis perfective, realis perfective and imperfective aspects.
The distinction between the perfectives is nearly always just tonal: there is no rule that accounts for the different final tones across all verbs, and a total of zero (segmental) suffixes to mark this distinction.
There are just two perfective suffixes, –l and –n, and just four imperfective, -n, -d, -g, -s (which actually surfaces as -i; long story); an imperfective suffix –l appears in just one verb. In addition, many verbs take no suffixes in one or both aspects.
Fusion doesn’t really happen. The only case where the suffixes are not immediately recognisable is that perfective -l appears as -d after originally monomoraic verb stems.
Unfortunately, which suffixes (if any) any particular verb will take is hard to predict. There is absolutely nothing, for example, about the form of lol ‘tie’ to tell you that its imperfective is lolg whereas the imperfective of bul ‘sow’ is buul.
The system is not quite as arbitrary as it initially appears, but the rules predicting conjugational forms are so complex (and have so many exceptions) that as a foreign learner your best bet is really to memorise the forms of each new verb you encounter separately. The tones of the various forms are only partly predictable, and partly depend on whether the verb in question has any derivational suffixes (which is not always apparent, as those are subject to fusion.)
This means that despite having a grand total of three flexional forms and hardly any fusion in conjugation, Moba verbs impose quite a big load on the memory of any learner. By any sensible definition, the system is very complex.
The thing is, Moba is not unusual as an example of this kind of complexity (I only thought of it first because I’ve recently been trying to account for some of the complexity on historical lines.)
Coming up with sensible metrics of “exotericity” is no easier, though you can’t say that they are unaware of the problems, which they do try to address by trying out several different criteria.
Overall, this seems to suffer from all the usual problems. As they used to say in the intro to Slashdot polls. “if you’re using these numbers for anything serious, you’re insane.”
…yabbut… how long has it been that a large number of non-native speakers has been learning Standard German – from professional teachers with textbooks & stuff? No dialect has the full four-case system, for example.
That said, the two adjective declensions and the three largely unpredictable genders hardly seem to be going anywhere.
They figured out an answer, but they admit the question might not be the right one.
No dialect has the full four-case system, for example.
If one wanted to learn quick-and-dirty spoken Standard German, not totally grammatical but serviceable everywhere, which case would be most easily jettisoned?
Genitive. Nobody needs genitives.
they admit the question might not be the right one
I’ve gradually learnt in my own work that asking the right questions is half the battle. In fact, finding the right question is often significant progress, even when you don’t have any idea of the answer.
What’s going on with their Figure 3: “The scores of fusion and informativity on the global tree”?
What “global tree”?
The caption witters about ancestors and descendents and phylogenetics. Please tell me that they don’t imagine that they’ve managed to create a tree of all the branches of Proto-World. (I note that “Atlantic-Congo” is a sister to all the rest. KONGO!)
I may have misinterpreted them. But if not, this looks like sheer Ruhlen-level pseudoscience.
German is learned and spoken by a large number of non-native speakers, and yet, it has retained its case system and many other grammatical distinctions.
The reason, I suggest, is this: native speakers don’t take non-native speakers that seriously. I myself am certainly not inclined to “jettison” the case system, and other “distinctions”, merely because a bunch of furriners struggle with it. Can you imagine the French doing such a thing ? If Houellebecq’s novel Soumission had been about language, it would have been so implausible that nobody would have bought it.
The very fact that (some) furriners do indeed struggle with German, instead of giving up, deserves respect. Imagine their dismay at reaching the heights, only to find Germans jettisoning it all in a fit of linguistic Nächstenliebe.
If you don’t write too much, it’s going to take a long time till anyone even notices you don’t know the genitive.
The idea is that if lots of people immigrate and never learn the language properly, the next generation will end up speaking it without some of the rarer irregularities.
What“global tree”?
Ref. 44, here. The procedure for attaching family groups into a higher order tree is explained beginning on p. 29, especially p. 31 (phylogeographic model).
It’s not Grampa Ruhlen’s snake oil. It’s the best, State-of-the-Art snake oil.
I could say something juvenile about “posterior trees”, but that would be beneath me.
…It says “phylogenetic” in the figure caption. 😮
Searching the paper for “global tree” finds it’s from reference 44, which is this preprint. And that contains a…
…supertree. That’s a tree made not directly from data, but from published trees that are used to fill in each other’s gaps. This is interpolation, not extrapolation, so there must be a Ruhlen-type tree of the major language families of the world in there.
What a searing failure of peer review.
Edit: Reading what seems to be the relevant part of the preprint heavily implies it’s not a supertree from published sources alone; instead, the backbone comes from an actual Bayesian phylogenetic analysis, I have not found of what data, repeated 27 times so you get a wonderful Bayesian probability cloud. Then the 27 published trees, each of one of 27 language families, were plugged into each of the 27 backbone clouds. ~:-|
Anyway, the preprint emphasizes: “Crucially, our findings are not contingent on a single tree of world languages, but rather integrate over a range of possible alternatives given the considerable phylogenetic uncertainty inherent in our model and priors.” Apparently the authors of the Science Advances to the Rear paper didn’t even understand that much, just took some average of the probability cloud of probability clouds and called it “the global tree”.
The idea is that if lots of people immigrate and never learn the language properly, the next generation will end up speaking it without some of the rarer irregularities.
Haven’t seen that happen with the Turks here. Those who never learned German properly go back to the Homeland when they retire, unless they become taxi drivers.
Young Turks nowadays speak German just fine. ‘Twas not ever thus. Any peculiarities were appropriated by Was guckst du? and benevolently chuckled over. The effect of that on standard German has been zero.
Yes. In situations with more massive immigration and less education, outcomes can be different.
Outcomes are often differenter on the other side of the fence.
Ref. 44, here
Good grief, what a parody of science. I note some familiar names among the perpetrators.
Ruhlen (following Greenberg) used imaginary linguistic data. These guys use non-linguistic data, which is merely arguably imaginary.
Evidently historical linguistics could be made much more scientific by eliminating the linguistics and the history, both of which are notoriously subjective and error-prone as disciplines (if you can even call them that.)
I wonder if the paper was in fact written by ChatGPT?
Oh yeah, the preprint tried to constrain a few nodes from non-linguistic data: monogenesis in Africa a reasonable time ago, single exit from Africa a reasonable time ago, single entry into the Americas for no good reason whatsoever while airily dismissing the obvious counterarguments…
If you don’t have flour, but really insist on baking bread, use sand. Tell everyone it’s the best sand available.
If one wanted to learn quick-and-dirty spoken Standard German, not totally grammatical but serviceable everywhere, which case would be most easily jettisoned?
All of them: the result would not be Deutsch but Dutch. The dialect spoken across the Channel is of course Double Dutch.
Genitive. Nobody needs genitives.
Except English and Swedish speakers. Most Germanic speakers decided the genitive was the only case worth keeping. Typical Deutsch to choose accusations over the warmth of procreation.
But they still like giving.
Obviously Swedish-speakers have long lived in closer proximity to other nations than Icelandic-speakers, but I had thought that Sweden having a significant percentage of residents who were not L1 Swedish-speakers and/or who were Swedish-born children of immigrant parents with a non-Swedish L1 was (modulo some Hanseatic German traders in the very old days) a quite recent development of the last half-century or so that cannot be a causal factor in syntactic changes long predating the recent rise of mass immigration.
Although speaking of mass immigration, Iceland’s current population is approaching 25% foreign-born, but this is an even more recent phenomenon (since the 1990’s). So stay tuned for any syntactic impact on Icelandic, although apparently the single most common immigrant L1 is Polish, which is not a language with an impoverished case system.
not a language with an impoverished case system
The people who actually subscribe to these sort of “explanations” for language simplification are long on just-so stories rather than anything amenable to verification or refutation, but FWIW they often claim that interference from a second language which has similar grammatical categories but a different way of marking them morphologically may lead to morphological simplification.
For all I know, they may be right. Western Oti-Volta has an agreeably simple system of verb conjugation compared with almost any other Oti-Volta group, and WOV has almost certainly spread over its current area (four times the size of Wales!) fairly recently. The most likely candidates for substrate languages are Grusi languages, which have pretty much identical verb categories to WOV but mark them very differently.
On the other hand, maybe the Mossi-Dagomba warriors just said to themselves: “Bugger all this morphological complexity! We’ve got people to conquer here and no time to bother with all that namby-pamby fusion and suppletion! That’s for girls!”
We had a dialogue — oh, ages ago — about Korean (and Latin/Greek) losing phonemic vowel length — and went on to consider other phonemic change resulting in reduction in total number of vowels/phonemic distinctions.
Would these Max Planck geezers include that metric in ‘complexity’? Loss of a phonemic distinction might lead to collapse of a grammatical marking? (Or will a language not lose a phonemic distinction if that would be the result?)
Why not include phonemic complexity in the overall measure of complexity? And the just-so story about social environments.
Discuss wrt PIE and its small-ish phoneme inventory.
Phonemic complexity is not part of what this paper is considering, perhaps because the idea is to showcase the wonderfulness of Grambank rather than WALS.
Swahili-as-lingua-franca, which presumably counts as more exotericker than yer actual coastal Swahili, has apparently lost the distinction between aspirated and unaspirated voiceless stops (which “proper” Swahili has, though it’s not marked in the orthography.)
PIE had 24 consonant phonemes (I think), which is not particularly few (more than English.)
Proto-Bantu, now: that‘s pretty short on consonants (eleven.) More vowels than PIE, though (seven.) And both had tones, though proto-Bantu had a more complicated tone system than PIE overall.
“Imperial” Western Oti-Volta (i.e. not Boulba, away off in Benin not conquering anybody) has lost phonemic distinctions present in most of the rest of Oti-Volta, notably having lost all the palatal stops. On the other hand, it’s kept some distinctions mostly lost elsewhere: the phonology doesn’t give any overall impression of simplification, unlike the morphology, with its nice simple extremely regular verb conjugation, near-complete loss of two of the original ten main genders, and abandonment of all agreement by gender.
I’ve read a whole list of scientific papers in several fields that were like that.
That does happen sometimes. The example that comes to mind is the loss of the instrumental case from OHG to MHG: its ending was a vowel, and the distinctions among unstressed vowels collapsed between OHG and MHG. Poof, the case was gone. (And we’re still not quite over it. English AFAIK lost it the same way and basically replaced it with by, mostly; in German it’s mit, but sometimes durch is preferred, and sometimes everything sounds off.)
3 × 5 − 1 plosives, 4 fricatives, 2 nasals, 2 liquids, 2 glides – yep, 24. Though it looks like at least one labialized fricative is coming.
(*[b] existed, but only in */pd/ and */pʁ/.)
PIE did not have tones; the tones of Classical Greek and Late Vedic are transparently late and independent developments, the Balto-Slavic tones can be blamed on laryngeals and Winter’s law (in a way actually quite similar to the Greek and the Central Franconian ones), and the Proto-Germanic distinction in some word-final vowels that was once thought to be tonal and is nowadays often represented as long vs. overlong vowels is probably just long vowels vs. vowel clusters.
What about Verner’s Law?
It occurs to me that we’re using “tone” in different senses, and certainly the proto-Bantu every mora-is-high-or-low-tone system isn’t like anything I’ve ever seen proposed for PIE: that’s what I was attempting to say by “more complicated.”
Presumably you mean that PIE had emic stress rather than tone?
(Proto-Bantu didn’t have contrastive stress, of course. Not many Volta-Congo languages do. So another point for PIE “complexity” …)
You say “pitch accent”, I say “tone …”
https://en.wikipedia.org/wiki/Proto-Indo-European_accent
It just occurred to me that I can’t think of all that many top-level protolanguages which do have tone in the baked-in pervasive manner of proto-Volta-Congo.
Isn’t it inappropriate to compare the evolution of languages in the times of schooling and specialized language education, vs. the era of marketplace pidgins?
I think the authors try to deal with that by playing with various different measures of exocitity/exoticity/whatever, but that’s the sort of thing I had in mind when saying that defining that is no easier than defining “complexity.”
There are lots of very different ways of being multilingual, and more than one way that a language can influence or completely replace another. Trying to assign a single number to them and and expecting to do meaningful statistics with it all strikes me as a mug’s game, even if you’re prepared to adopt the unprincipled approach of fiddling with your definitions until you get the answer you want.
I think this is the WALS/Grambank problem in a nutshell: the homogenisation process needed to fit all these very different languages into the Prucrustean bed of a single overarching labelling scheme destroys information, and not marginally or trivially but extensively. No amount of fancy statistics can bring the information back again. If they could do that, they should be bending their efforts toward getting a perpetual motion machine going and solving the world’s energy problems.
Ah, but PIE did not have pitch accent. All the introductions keep saying it did, but they’re simply not using the term in its usual meaning. Instead, they’re trying to say – sometimes explicitly saying – that the single stress per word-or-so was “chiefly” realized as higher pitch “rather than” loudness. That’s a mixture of trying to say “mora-timed as opposed to stress-timed” and of the old attempts to reconstruct a (very limited) pitch-accent system to account for the phenomena I listed and dismissed. For the last few decades there’s been a consensus that PIE had only one kind of stress, not two (high and low, “acute” and “circumflex”, or whatever).
“Is it true that Comrade X was in Y and did Z?”
“Yes, except it was Comrade A in B doing C…”
“Is it true that PIE had pitch accent?”
“Yes, except…”
It is clear that the position of the stress was phonemic and not limited to particular positions in a word; very much like Russian, or Modern Greek minus the restriction to the last three syllables, or Lithuanian minus the tones.
Several others that could have had it haven’t been reconstructed far enough. Tone has a long history of being ignored by Caucasianists, and once the full horror of Proto-Oto-Manguean will be understood, we shall either go mad from the revelation or…
On Verner’s law specifically: I think it’s a general gonsonand weagening that applied everywhere except word-initially, right after the stress, or to consonant clusters – precisely the “positions of strength” where consonants are most likely, or are most likely reinterpreted as, long or aspirated. Right after the stress is the only position where my kinds of German allow consonant length, for example; and in the High German consonant shift, the plosives in word-initial position and in clusters, where the length distinction didn’t apply, all behaved as long.
@DE: To play devil’s advocate, you can say that science is a never-ending sequence of approximations, and that describing multilingualism by a number is a better approximation than ignoring it. The question is whether this is indeed an approximation, or a qualitative mis-description that throws the whole model off.
This is comparable to arguments for and against tree models in general: are they universally useful as a first approximation, to be refined with cross-links representing contact? Or should they be discarded in favor of something else?
Several others that could have had it haven’t been reconstructed far enough.
The various Khoesan families, too.
@Y:
Yes, the choice may in reality be between doing it badly and not doing it at all, and there is always the consideration that doing it badly might provoke someone to do it better.
As with Mass Comparison, the technique itself is not irredeemably evil or anything: it’s a question of how you interpret your findings and what that leads you to do next. The real mischief comes when you (or others) think you’ve already solved your problem and future research is unnecessary, or when overinterpretation of your results sends future research off on wild goose chases. (This has happened with Greenberg’s “Niger-Congo”, for example, or Gabriel Manessy’s “Gur”, by diverting people away from patient bottom-up reconstruction instead of encouraging it. Rather like researching Nostratic before you’ve actually established PIE.)
The various Khoesan families, too.
True. I was sneakily trying to armour my comment against refutation by smuggling in the helpfully vague “top-level”, but you’ve trapped me: I could only disagree by saying that “Khoisan” as a whole is a valid group. Well played!
(Good point about inadequate reconstruction, too, DM.)
general gonsonand weagening that applied everywhere except word-initially, right after the stress, or to consonant clusters – precisely the “positions of strength” where consonants are most likely, or are most likely reinterpreted as, long or aspirated
Actually, that all applies pretty closely to proto-Oti-Volta too. The only positions where voicing contrasts were possible were root-initially (usually the same thing as word-initially) and (probably) directly after a root vowel. The evidence is a bit open to interpretation, but it seems likely that everywhere else, stops were always voiced. And voiceless stops after root vowels may in fact all derive from previous consonant clusters, too.
Yes, except…”
I think you mean “Ye-es, except…”
I’ve never heard a Radio Yerevan joke.
‘they often claim that interference from a second language which has similar grammatical categories but a different way of marking them morphologically may lead to morphological simplification’
Seems like a reasonable idea in explaining Middle English (as a result of Norse influence), but it’s a hard kind of thing to be sure about.
‘The example that comes to mind is the loss of the instrumental case from OHG to MHG’
This was well on its decline before the reduction to schwa, getting replaced by the dative in the 10th century. This followed a period where bare instrumentals were replaced by prepositional phrases, which may have been an enabler for the loss of the increasingly non-load-bearing instrumental.
‘probably just long vowels vs. vowel clusters’
This has some difficulties, namely that some ‘circumflex’ vowels come from simple, unclustered vowels etymologically. I’d prefer to say (and do say in my book) that we just don’t know what kind of contrast was happening here.
Oh, interesting, so it was a der Dativ ist dem-Genitiv-sein Tod situation…
…So I finally looked you up in Google Scholar, and found lots of interesting stuff, but no book. Is your book out yet?
The book came out in April, but I guess it takes some time for Google Scholar to catch on? It’s open-access here: https://global.oup.com/academic/product/prosody-in-medieval-english-and-norse-9780197267462
I’m just a layperson and can’t back this up with scientific data, so take with caution. However, regarding German, I’d say that, performance-wise, even native speakers (not writers) get the case endings wrong all the time. In theory, this should produce all kinds of misunderstandings, but here’s why it actually doesn’t:
– While in theory, word order is fairly free, it is in practice much less diverse than theoretically possible.
– The case endings are hardly distinguishable any more (lots of overlap, and all vowels are schwas), so they wouldn’t be reliable under less-than-optimal circumstances anyway.
– Speakers re-structure their sentences mid-utterance. You start with a dative, but when you’ve finished talking, it should in retrospect have been an accusative because you chose a different verb which reigns a different case, or whatever. Do you start again and correct yourself? Hardly, since you got yourself understood regardless.
This is, by the way, why I imagine noun-phrase declension might be one of the hardest parts of learning German as a foreign language: On the one hand, it’s quite a complex affair (in short, it’s not just arbitrary grammatical gender, but the noun phrase as a whole is declined differently depending on determinacy [is that a word? you get my meaning]), but on the other hand it bears almost zero functional load any more—so, meager rewards for a considerable learning effort.
I presume the situation to be different among speakers of languages where case-endings are indeed distinctive and important for unterstanding. (Icelandic? Polish? Lithuanian?)
In written German, of course, wrong case endings do get noticed and are socially sanctioned against. This is certainly one of the reasons why second-language learners, possibly including L1 speakers of divergent dialects, and, in general, people who for whatever reason didn’t get an extended education in written standard German, while they may be perfectly functioning in spoken German, are mislabelled as dumb and denied opportunities in contexts which demand writing prose. (Perhaps there’s a somewhat comparable situation in English-speaking societies regarding mastery, or lack thereof, of the arcane spelling rules, I don’t know.)
I imagine noun-phrase declension might be one of the hardest parts of learning German as a foreign language
The _idea_ of case/declensions and agreement across a whole NP was familiar enough. (I’d learnt Latin at school.) It was the sheer rote memory of which noun was which gender; and which verb governed which case. At least with English it’s a relatively small number of (strong) verbs that form irregular participles, and they’re frequent so memory gets reinforced.
Perhaps there’s a somewhat comparable situation in English-speaking societies regarding mastery, or lack thereof, of the arcane spelling rules,
Yes that’s exactly how it is. We have the ‘greengrocers’ apostrophe’ as stigmatising those who can’t (be bothered to) tell plurals and genitives; we have special homophones (right, write, rite; read, reed; read (pp), red) whose differentiated spelling is entirely arbitrary. And spell-checkers are barely less arbitrary in guessing which one you meant.
Classical Arabic is a good example of a language with an elaborate case system which actually has an extremely low functional load. You have to really work at it to contrive sentences which become ambiguous when you drop the case endings. (Just as well, because otherwise written Arabic would become a morass of doubletakes and garden-path sentences,)
The sub-Chomskyan idea that languages are somehow always “efficient” is basically just false. In fact, redundancy is a key part of how reliable communication works.
>The _idea_ of case/declensions and agreement across a whole NP was familiar enough. (I’d learnt Latin at school.)
Indeed, but I believe stuff like
das blaue Haus (‘the blue house’)
ein blaues Haus (‘a blue house’)
(both nom. sg. neutr., but note the different endings) or
der gute Freund (‘the good friend’)
mein guter Freund (‘my good friend’)
(both nom. sg. masc.)
is a Germanic specialty. (Dutch is similar in principle–het blauwe huis, een blauw huis–but less complicated …*I think.* My Dutch is not good.)
>Classical Arabic is a good example of a language with an elaborate case system which actually has an extremely low functional load.
But those endings are at least audibly distinct? The inventory of case endings in spoken standard German is, if I’m not mistaken, [n̩], [m̩], [ə], [ɐ] and [əs].
To Stu Clayton:
>Young Turks nowadays speak German just fine. ‘Twas not ever thus. Any peculiarities were appropriated by Was guckst du? and benevolently chuckled over. The effect of that on standard German has been zero.
I’m not so sure. I lack the education and vocabulary to describe it adequately, but I notice a shift in accent/speech rhythm/timing/??? in younger people whom I presume to be native speakers, including German monolinguals. I’m hard-pressed to convey what I mean, but I’d say the way many young, urban (?) people talk is at the same time ‘choppier’ and ‘slurred’, although that may seem contradictory. And, back to the point, I’m not convinced why L2-influence in the wake of considerable immigration should be excluded as a possible (partial) explanation.
IIRC the mein guter Freund/der gute Freund mixed-declension thing is uniquely German, rather than Germanic: and modern German at that (not there in Middle High German, which just had strong and weak adjective declensions, like any respectable Germanic language.)
My grandfather was once accused of speaking German too well, by which the accuser seems to have meant that he got all the endings right all the time. I’m told that German proofreaders are accustomed to seeing multiple errors, especially in adjective declension, in texts written by perfectly competent native German speakers. Ordinary readers tend not even to notice them.
PS, to Stu Clayton:
>benevolently chuckled over
Chuckled over, yes. Benevolently? I envy your positive outlook on society.
To David Eddyshaw:
> like any respectable Germanic language
Do respect the Dutch! Otherwise they’ll sell you an umbrella and laugh at your being swept away by a rainy North-Sea breeze …
All the Dutch people I’ve ever worked with (a surprising number, come to think of it) have been perfectly secure in their national identity and language (if somewhat prone to subtle mockery of Germans), so I’m probably safe.
However, Dutch does distinguish strong and weak adjective forms:
een klein huis
het kleine huis
(Even English did, as late as Chaucer’s day.)
The ‘das blaue Haus~ein blaues Haus’ thing is not particularly new in German, and I’m not sure which aspect of it is meant to be anything but a continuation of the normal strong/weak adjective pattern in Germanic, which is also found, in a more reduced form (except in Icelandic and Faroese), in the major North Germanic languages today.
Is it maybe the reduction of marking on the indefinite article (and similar), ein for einaz, einer, that’s meant to be of interest here, producing the ‘conservation of marking’ effect where an overt ending like -s or -er only appears once per phrase? That was already normally the case in Old High German, where the ordinary versions of these would have been daz blāuua hūs and ein blāuuaz hūs (or ein blāo hūs; there were competing forms for the neuter), respectively, at least in the nominative-accusative singular.
The weird thing about the German examples (writing as a native speaker of a North Germanic language) is that mein doesn’t trigger weak declension as a properly educated determiner should. I’m pretty sure we were larnt only the the strong and saucepan declensions for adjectives, but declining strongly after a possessive pronoun is positively unGermanic. If you ask me.
And yes, the ON determiner hinn was copiously inflected. The ones in current use in Scandinavia were stolen from the Hansa at a time when flexion was not as much de rigeur as earlier; hin/hint/hine can still be used, but only if you want to sound like Kierkegaard. Already then they had nothing left but gender.
>The ‘das blaue Haus~ein blaues Haus’ thing is not particularly new in German, and I’m not sure which aspect of it is meant to be anything but a continuation of the normal strong/weak adjective pattern in Germanic
Diachronically, yes. Synchronically, it’s just clutter.
To Lars Mathiesen:
I’m positive that I’ve been routinely committing errors like *’mit stort hus’ when talking Scandinavian. I have never learnt Danish properly. 🙂
I mean, to some extent the whole thing is kind of clutter all the way back! Even in a nicely behaved older Germanic language like Gothic, it’s just a bit weird to have one set of endings (based on n-stem endings) used in a slightly arbitrary set of syntactic contexts (definiteness is at least not the whole answer), and another, partly based on a- and ō-stem noun endings, partly on pronominal endings, used elsewhere. But weird or not, basic elements of the system survive in some form even in quite a few languages where the nominal system has been otherwise reduced to the bone.
As Lars says, what makes German adjective declensions unusual is that there are three patterns that have to be memorized, depending on the kind of determiner used. There’s a weak pattern for after der words, mixed for after ein words, and strong when no determiner is present. Of these, the strong is the least used, but ironically also the simplest, since the case endings are just the same as for the der words, except it has “-en” instead of “-es” in the genitive. But who uses genetives without a determiner anyway?
…and preceding labial consonants routinely turn the former into the latter, eliminating the acc.sg-dat.sg or the dat.sg-dat.pl distinction in that context.
Other than that, I’ve not noticed much chaos. But the rarest of the three (!) options, the singular without article (telegram-style/headlinese; absent in spontaneous speech), does seemingly trip most people up. A few years ago, Firefox changed the correct In neuem privatem Fenster öffnen into the wrong In neuem privaten Fenster öffnen. Irks me every time I right-click on a link.
The determiners themselves are in that position… and dieses Jahres is being replaced by diesen Jahres in large parts of Germany.
Tangentially relevant:
https://scilogs.spektrum.de/sprachlog/kein-bock-auf-nen-interessantes-grammatisches-ph-nomen/
For the record, while I follow the reasoning from a descriptive-linguistics perspective, this usage feels just WRONG to me. My idiolect: ein = ‘n; einen = ‘n, optionally ‘nen; einem = ‘nem, einer = ‘ner; eines = eines.
(Just to be clear, I do emphatically distance myself from the likes of Mr Sick.)
‘what makes German adjective declensions unusual is that there are three patterns that have to be memorized’
Indeed, and I guess that’s where diachronic thinking does mislead — I tend to think of all this in historical terms, as only two groups of endings, strong and weak, with adjectives after articles showing drift from the strong to the weak forms for the ‘peripheral’ parts of the paradigm (the plural and the gentive and dative).
That’s a regional phenomenon, the wholesale replacement of [n̩] by [nɛn] – it’s going to be a classic case of a language change that makes things easier for the listener instead of the speaker. Confused me when I first encountered it; it seemed like a replacement of ein by einen.
what makes German adjective declensions unusual is that there are three patterns that have to be memorized
Unfug. I never had to “memorize patterns”, if “patterns” means rules explicitly formulated in a book of grammar. ‘n Scheiß muß ich, as they say, and mußte.
I simply learned over time how this stuff works. Learning always involves memory, believe it or not – and practice too, and being corrected and learning to correct yourself. Nothing to see here.
One can shed tears over these oh-so-conundrous idols of the study (of German, or any other language). But that’s merely when one is on the outside looking in, like The Little Match Girl. Inside it’s party time.
If you don’t have the knack, lay off the bitchin’.
Of course I still trip up from time to time, in English and German, and am then covered with shame. Big deal. Shame is a great incentive to crank up the old attentiveness. Fact !
The methodology of linguistic description has little to do with the methodology of speech performance and language learning.
It means rules that could be explicitly formulated in a book of grammar. That includes rules that were never in fact written down, and even rules that nobody is consciously aware of.
>That’s a regional phenomenon
Spreading, I guess. And I hate it. Know a person who graduated in Frankfurt/M., incessant nen-user. Unfortunately, there’s several universities in Hesse …
>a language change that makes things easier for the listener instead of the speaker
Er, rather the other way around, isn’t it? Speaker has one less distinction to make, listener gets more ambiguity.
I mean specifically the replacement of [n̩] by [nɛn]. The earlier merger of [nən] into [n̩] – “earlier” in the sense that it’s a necessary prerequisite; it’s not completed – is indeed a change that makes things easier for the speaker and harder for the listener.
That includes rules that were never in fact written down, and even rules that nobody is consciously aware of.
Ah well, I guess you were born too late to have imbibed The Concept Of Mind along with mother’s milk. [I was born in the year it appeared.]
Ryle writes:
#”To be intelligent is not merely to satisfy criteria, but to apply them; to regulate one’s actions and not merely to be well-regulated. A person’s performance is described as careful or skillful, if in his operations he is ready to detect and correct lapses, to repeat and improve upon successes, to profit from the examples of others and so forth. He applies criteria in performing critically, that is, in trying to get things right” (pp. 28-29).
…
As one of Ryle’s students observes, “it is not repeated behaviors but changing behaviors that are the sign of creative intelligence, and this can only be observed in the long term” [Brainchildren, Dennett 1983]
#
That is, to get things right by one’s current lights. Doing this can turn the light dimmer up or down, for the next time. That’s how things work with me, at any rate.
This is a process, but not the kind performed by a ghost in a Turing machine.
At the same link.
#
Wittgenstein once said that only two people understood his philosophy, and that one of them was Gilbert Ryle.
#
Ah well, I guess you were born too late to have imbibed The Concept Of Mind along with mother’s milk.
Or, alternatively, like most of us here he’s more concerned with linguistic facts than with philosophy.
Or, alternatively, like most of us here he’s more concerned with linguistic facts than with philosophy.
Name-calling !
It’s good ideas that are needed here, under whatever names. They are available at every good bookstore and blog. Ryle, Wittgenstein, Dennett – not too much to expect, I think.
I never said language was ipso facto “creative intelligence” or a “careful or skillful” “performance”.
That’s my point. There is no obligation to choose only one set of ideas, to the exclusion of all others.
I found Wittgenstein tedious and banal at best (and irritatingly unwilling to do more than just Ponder–which I think is a little bit ironic). But plenty of people seem to like him, so maybe there’s something buried in there after all.
To David Marjanović:
But the replacement of [n̩] by [nɛn] eliminates a distinction the speaker previously had to make; the result of the replacement is a merger of the two into [nɛn] …
But plenty of people seem to like [W.], so maybe there’s something buried in there after all.
Well, the tractatus is certainly tedious and banal. I once read the Philosophical Investigations, without profit. I think DE is a devotee, I’m certainly not.
I mentioned Wittgenstein’s take on Ryle only because it was new to me, and might encourage W. fans to take a look at Ryle. Sorta like using sow pheromones to attract bores. [Actually, I think in real life it’s the other way around, the boar being responsible for inducing Duldungsstarre in the sow.]
I think the merger of [nən] into [n̩] came first, and then some of the people who had the merger replaced [n̩], regardless of its source, by [nɛn] – interestingly not [nən].
(I find this noticeable because I’m from too far southeast to have [ə] natively.)
I think DE is a devotee
Yup.
I think the trouble with appreciating Philosophical Investigations is that it’s a bit like coming into a movie halfway through. Unless you already have some idea why Witters seems to be making a big deal over some apparently bizarre and pointless issue, it’s difficult to see why he’s bothering (and why you should, either.) He assumes a lot of background* and doesn’t believe in helping you by spelling it out.
I don’t think his reputation is helped by some of his fans, whose enthusiasm for his works seems to be rather disconnected from their actual content.** The philosophy-rockstar thing is bound to put people off.
* Some of it, frankly, on a “What have I got in my pocket?” level. (No wonder Gollum was pissed off.)
** Like Jane Austen. Exactly like Jane Austen.
Wittgenstein did more than Ponder, he notably and even for the times excessively beat slow-learning children in various village schools (dunno why they kept hiring him, maybe they had suffered blows to the head as small children). But he later decided this was not a Good Thing and went back to Pondering…
** Like Jane Austen. Exactly like Jane Austen.
There have been rumors of a lost novel by him, tractatus superbiae et praejudicii, loosely based on his own life in Cambridge.
https://en.wikipedia.org/wiki/Sense_and_Sensibilia_%28Austin_book%29
‘Sense and Sensibilia’ is a really delightful title!
Pleasingly, it’s actually a perfectly apt description of what the book is about. It would have been an entirely reasonable title even if it hadn’t been so obviously necessary.
I was reading somewhere that where W crossed the line in his schoolteachery was not the corporal punishment aspect itself, which was entirely in accord with contemporary local norms, but in treating the girls just the same as the boys, on the grounds that they were every bit as capable of learning maths if they could be bothered/motivated. This was the cue for the pitchforks and flaming torches. Some things are just beyond the pale.
W was a hospital porter at one point (I actually worked in the hospital in question, though for some reason I don’t remember him.)
Sense and Sensibilia
Coincidentally or not, the author is named J. L. Austin.
Not.
[Actually, I think in real life it’s the other way around, the boar being responsible for inducing Duldungsstarre in the sow.]
In America at least, boar saliva can be bought commercially for this purpose. Just think, I could have been a boar saliva harvester.
@Rodger C: I could have been a boar saliva harvester.
You’re located near West Virginia or in the general Hannibal-the-film area, right ?
I couldn’t find a word or brief English expression corresponding to the fabulously succint Duldungsstarre, one of my favorite words. Not being a pig farmer, I don’t actually know the phenomenon. It just makes metaphorical sense in the context of you-know-what, even without chemical tricks. Duldung is quite sufficient in the absence of marital intent.
“Born in West Virginia, he began his career as a humble boar saliva harvester.”
I have, in fact, written a novel mentioning an Eastern European anti-Semitic agitator of the 1930s named Prince Pstislav Borsalaiva.
Wasn’t there a news story recently about the controversy over whether the former Borsalaiva family castle that had been confiscated in the late Forties should be returned to the emigre heirs by the post-Communist government, especially now that the Institute for the Study of the Application of Scientific Materialism to Podiatry that it had housed has gone out of business?
You shouldn’t mess with the Podiatrists. A friend of mine still bears the psychic scars of trying to get access to an outlying operating theatre in order to treat blinding eye conditions. The Pod People decided that they wanted the operating theater instead, and they won. They have Friends in High Places. (But I have said too much.)
@David E.: The dreary Communist-era Institute has been replaced by a more contextually appropriate successor (rumored to have been funded by George Soros) known as the Institute for the Study of the Application of Post-Privatization Oligarchic Capital Flows to Podiatry. It disdained the drafty old castle and is housed in a fancy new building of breathtaking ugliness that a Prominent International Architect received some astonishing number of Euros to design.
#
Saint Pstislav of Perm cured a beggar of leprosy. … He was pronounced a German and deposed …
# [And Quiet Flows The Vodka: Or When Pushkin Comes To Shove: The Curmudgeon’s Guide To Russian Literature With The Devil’s Dictionary Of Received Ideas]
I still don’t know how to pronounce the name, German or not. Do you slip an almost inaudible “e” in between “P” and “st” ??
“Τηισ παγε ιντεντιοναλλψ λεφτ blank”
No. You say “Pssst!” and then, instead of heeding your own advice, you suddenly begin to speak.
Slavic shenanigans.
From the novel, set in 1979, overheard on the radio: “Then it’s to Castle Borsalaiva, now the National Palace of the People’s Heritage, where once upon a time the feast of the Virgin of Ropbgba was celebrated with a roast aurochs stuffed with swans!”
In a sequel, set in 2008, the son of a Jewish refugee is visiting: “Castle Borsalaiva. The squat gray shape seemed to breathe a cold air down on him from his father’s stories. And so be it. It wasn’t even the People’s Palace of Whatever any more; in fact his guidebook didn’t mention it at all. He strongly suspected it was a CIA site. At that thought, the cold wind that hit his bones but not his skin changed its smell but not its coldness.” This is shortly before he flees the village at the approach, not (like his father) of a column of horsemen, but of a Heritage Tour bus. Later he discovers that the whole place he visited was a rebuilt “model village,” the original site being in the middle of an open-pit copper mine. I needed you guys around when I was writing all this.
Re “Ropbgba,” before the question is asked: “The first thing Jay had noticed was that the river he’d always called the Wlw was called the “Wlrwl” on the guidebook map. What on earth. Anyhow the Borzokian language wasn’t Slavic, as he’d always assumed; the guidebook said it had evidently arrived with the Avars, fifteen hundred years earlier, from somewhere in Siberia, and had no certain relatives. Nevertheless it was full of Slavic words and was part of a Slavic-dominated Sprachbund, which sounded to him like some sort of Yiddish organization.”
You know, people talk reverently about roast aurochs stuffed with swans, but I personally find the swans detract from the flavor — give me plain roast aurochs any day.
They have Friends in High Places
Quite the opposite, for the podiatrists. You are thinking about the scalp specialists.
On second thought, I take it back. Podiatrists, like bootblacks, put their clients in high chairs.
“The results show that societies of strangers do not speak less complex languages.”
Wwwwwow!!
(ironical “wow”. Meaningless irony, I just find the fomulation funny)
How do you measure informativity based on a database? Assume, I don’t mark number. I can still retain informativity if
– unlike everyone else I mark gender (and – hypothetically, of course – it is not in the database, because people who complied it simply don’t know what’s “gender” and didn’t notice it).
– I mark number lexically (with the following discussion of whether isolating languages are “less complex” than synthetical langauges and whether I mark number by “grammatical” means of an isolating language or just lexically)
So how can we demonstrate that our database is “complete” in some way (or representative), and in what way?
I assume, a complete catalogue of word forms is indeed possible…
Every Cauchy sequence in the database has a limit?
Wwwwwow!!
Well, the contrary has often been asserted, so if they had actually demonstrated this rigorously it would be worthwhile. In fact, even if one indulges in some scepticism about their methods, they do provide some perfectly real evidence the other way, for all that it hardly settles the matter conclusively.
Personally, I’d tag Russian as a better individual counterexample than German, but that may just reflect the fact that I know very little Russian. Everything is complex if you don’t know much about it.
(Their own rather idiosyncratic way of measuring morphological complexity may make German come out more complex than Russian: I’m not sure if it would do justice to the unpredictable nature of Russian aspect marking, which is rather like my Moba example above: sure, there are rules and general tendences, but at the end of the day the poor foreign learner is just going to have to memorise a lot case by case. I don’t think it would capture the myriad difficulties occasioned by Russian stress changes in flexion, either.)
Kusaal has a regular distinction between main and subordinate clause verbs which is (a) completely orthogonal to tense, mood and aspect and (b) is usually marked exclusively by a tone overlay, but in certain cases is marked by an enclitic particle and in certain other cases by a distinctive (fused) flexion.
What complexity score does that get? Presumably zero, as the category in question doesn’t seem to figure in the database at all.
This is not an unusual case cross-linguistically. There are entire dimensions of complexity in every language which a one-size-fits-all questionnaire, however detailed, will inevitably miss completely.
In fact, the enterprise reminds me of the old Croom Helm grammars, in which every language is forced into the mould of the Lingua questionnaire, resulting in an entire series where skilled linguists are denied the opportunity to describe languages adequately. Dreadful waste of scholarship.(The more enterprising authors try to work around the damage by supplying lots of appendixes for all the stuff that doesn’t conform to the cookie-cutter method or presentation.)
Well, one can measure correlation between presence of a feature and any variable (like number of L2 speakers) and say: for every feature that we tried, we see no corrlation between its marking and our variable.
Indeed, one does not have to be able to compare complexity of two languages to do that.
True: but then you also can’t say whether “complexity” correlates with anything at all.
I suppose you could play about with your definitions of both “complexity” and “exoticity” until you did end up with some nice correlations (there are plenty of degrees of freedom there.*) Then you could retrospectively pretend that the very correlation showed that those were the “right” definitions all along, or (if you were a real villain) simply pretend that you had selected those as your criteria on quite other grounds before you ever ran your nice statistics program at all.
Happily, the authors seem to have resisted going down this primrose path to pseudoscientific perdition.
* Which means you will find correlations if you look hard enough: unfortunately, correlations found in such a way are of no significance whatsoever, and prove nothing at all.
“True: but then you also can’t say whether “complexity” correlates with anything at all.”
You can say something, like:
– no correlation is consistent with the hypothesis that “our variable is not associated with uniform decline in compexity. Either our variable has nothing to do with complexity, or the features in our sample are insensitive to our variable.”
– correlation is consistent with “either our variable associated with uniform decline in complexity, or the features in our sample are particularly sensitive to our variable.”
(All of this assuming that we managed to avoid circularity that you describe in your comment:))
P.S. Imortantly, the issue now is not eurocentricity of your sample of features.
Circularity: I recently walked in a forest when it was raining and heard in the noice of raindrops falling on leaves something like a voice of a woman yelling at a distance. No words, just timbre and rhythm. Yes, when you amplify signals, you find them in noice.
You say “Pssst!” and then, instead of heeding your own advice, you suddenly begin to speak.
“Pssst!” is not, at least in English, an injunction to silence; rather it is a way of surreptitiously drawing someone’s attention to oneself. So following it with speech, perhaps in a whisper, is perfectly reasonable.
Podiatrists, like bootblacks, put their clients in high chairs.
Mine puts me in a sort of simplified dentist’s chair, without the oral apparatus.
Everything is complex if you don’t know much about it.
Nnnnno. The Chomskyites think Language is simple because they know nothing about languages.
The more you know, the more you know that there are no simple answers. “How do you know you’re interviewing an expert? — All their answers start with ‘That depends’.” Politicians are the ones with the simple answers (and questions).
Next time someone says “voulez-vous coucher avec moi?” to me, I respond “there are no simple answers!”
(“Expert texpert choking smokers…”)
Presumably, the more you know about the person asking, the more likely you are to answer like this.
It is in German, so I’m sure Stu is familiar with it. I’m aware the English translation is Shh! – which is understood in German (of course you can also find Pscht!), but not often used.
When I worked in a thrift store in San Diego, a few miles from Mexico, I had a customer with very little English who’d cheerfully address me, not with “Hi” or “Hola,” but with “Psssst!” I figured that he’d learned his English in a context of surreptitiousness, i.e. that he didn’t have a green card.
In the south of Ghana, you hiss at people (waiters etc) to get their attention. Northern Ghanaians don’t do this, and seem to find it as off-putting as Brits do.
It is in German, so I’m sure Stu is familiar with it.
¡Por supuesto! I momentarily forgot about shhh. Context is all.
“Only contéxt.” I guess only old folks like Hat, DE, JWB will understand that.
If languages tend to simplify over time, eg loss of Indo-European inflections, why did they ever start complicated?
Languages don’t simply simplify over time. There are always sound changes that make simple things irregular, for example.
Also “start” is begging the question. It so happens that the oldest attested languages are pretty complicated, but it seems likely that if we could push our reconstructions a bit further back, all the synthetic inflexions would resolve into postpositions and clitics and then into indeclinable (“analytic”) particles.
Julian: languages do NOT simplify over time. Modern English is much more complicated than Old English in many ways: if you are a native speaker of the former its complexities are typically things you are (consciously) unaware of. But to a non-anglophone learner (especially to a speaker of a non-European language) neither would seem globally simpler than the other.
This point was driven home to me rather forcefully lately: I was consulting and comparing a grammar of Classical Armenian (which was spoken about fifteen centuries ago) and another of Modern (Eastern) Armenian, and while as a linguist I could have deduced (if I hadn’t known it already) that Modern Eastern Armenian is a later, changed form of Classical Armenian, I definitely cannot claim (based on the content of each grammar) that Modern Eastern Armenian is simpler than its Classical Armenian ancestor: indeed, my impression is that either form of Armenian -to a linguistic outsider such as myself, i.e. one who is not fluent in any Armenian variety- would be about as complex (i.e. difficult to learn) as the other.
… and then into indeclinable (“analytic”) particles
Like “ugh” and “eee”.
Old Irish is the classic example of how you can complexify your language enormously through perfectly regular sound changes.
But then, even French has irregular verbs which make Latin seem positively agglutinating in its morphological simplicity.
@Lars Mathiesen, even when postpositions and clitics are discernible, it does not guarrantee that the previous stage was isolating.
One measure of complexity in the post is fusion.
Julian asked about inflexions, and indeed, we can tell a rather detailed story of how these can be abandoned and replaced with analytical constructions, but we don’t have a good example of a language who has developed them over the period of attestation.
I find his question meaningful.
What if we measure change in analyticity over time for all langauges that we were able to observe for at least centuries? How many langauges will show growth (and how much of it)? How many will show decline and how much of it? What will be the average for all langauges we know?
Is growth associated with certain regions – or maybe sociolinguistical circumstances?
drasvi: but we don’t have a good example of a language who has developed them over the period of attestation.
French.
(Some phase of) Old Egyptian, from what I hear, but I can’t judge that.
Coptic seems to be about the only good real example of a language completing the supposed inflecting-analytic-inflecting cycle; though, to be fair, hardly any other language than Egyptian has a long enough documented history that you could tell. I suppose that modern Indo-Aryan languages have managed to recreate a pretty impressive case system, though it’s not a patch on Sanskrit for sheer complexity,
I agree that it’s a perfectly reasonable question how e.g. PIE could have got that way in the first place: it’s not just the fusion that needs explaining, but the way different stems often inflect with completely different affixes to express the exact same category. When people deign to treat this as a real question at all, their answers often seem quite hand-wave-y. The affixes must have developed from clitics or the like, because … they just must have, OK?
@Trond, what French forms we are talking about (I understand that conjugation, but what part of it)?
Anyway: Latin was not exaclty “more isolating” language. French can be instructive in that it shows how forms change, but it is not an “arc” of a hypothetical circle (by which I don’t mean there is such a cycle) where inflexions evolve in a formerly isolating langauge only to be lost later.
Ditammari seems to have developed noun-class prefixes out of proclitic demonstrative elements at least in part because regular sound changes have eroded away much of of the inherited noun-class suffix system.
A kind soul recently pointed me in the direction of the Miyobe language, spoken just south of Ditammari, which looks pretty typically “Gur” and is lexically close to Oti-Volta, but has no class suffixes, only prefixes: I suspect that it’s just taken the same process as in Ditammari to its logical conclusion.
More broadly, I think it is almost certain that Niger-Congo class affixes have everywhere developed from proclitics or enclitics; and in e.g. Yoruba the former noun class prefixes are now reduced to individually meaningless vowels which begin most noun stems and have no longer have any grammatical roles at all.
@drasvi: We’ve discussed it before. Two links from that thread:
Peter Arkadiev: Grammaticalization of polysynthesis (with special reference to Spoken French) (class handout, 2005)
Guillaume Jacques: LE FRANÇAIS PARLÉ, LANGUE POLYSYNTHÉTIQUE? (blog post, 2021)
Here’s a partial treatment of noun cases.
@DE it’s a perfectly reasonable question how e.g. PIE could have got that way in the first place: it’s not just the fusion that needs explaining, but the way different stems often inflect with completely different affixes to express the exact same category. When people deign to treat this as a real question at all,
I take DE to be asking (for example wrt Latin) why there are 5/6 +variations different declension models.
@DM Here’s a partial treatment of noun cases.
Thank you. This does consider DE’s a ‘real question’. It posits cases came from fusing postpositional enclitics with a noun stem. Then some sound changes at the junction points that had the effect of obscuring there was a common enclitic.
Earlier the paper has noted different cases (their fusion with enclitics) arrived at different points in time in (Pre-)PIE, so we shouldn’t expect a consistent pattern at the ‘junction points’.
I’m not convinced this is sufficient as an explanation. Some Latin declensions Dat., Abl. pl. end in -Vbus, where the V represents some morphable connecting vowel, whose exact form depends on the stem noun’s ending. Typically -ibus, but for example collapsed in rēbus. Some declensions end in -Vs, typically -īs.
So we’re left with -Vs being a collapsed form of -Vbus?, with only the -s as residue? What would be wrong with ?mensibus or ?dominibus that they’d want to collapse to mensīs, dominīs? (Of course I mean what would be wrong with whatever were the proto- forms of those.)
(And I tried to pick an example from my very rusty schoolboy Latin that I thought would give most support to the hypothesis.)
@Trond, thank you! Indeed, when one defines “word”, French naturally comes to mind.
The exact relationship between this and
(1) directionality of change
(2) polysynthesis
is unclear because we need to define 1 and 2 first.
@drasvi, (about particles), yes, it’s something that seems to be glossed over in textbook treatments: That even if we can reconstruct aspects of a protolanguage that must have been ancestral to things we find in attested descendant languages, the actual spoken language would have been full of irregular inflexions and suppletions and what have you, that have been lost, regularized or analogized away in all the descendants (or maybe in all but one which means that they are not available to the reconstruction process).
So indeed, even if later noun paradigms arose from combinations of roots and affixes, the Uniformity Principle tells us that it’s very likely that there were other ways of identifying verb arguments (for instance) in concrete language where those combinations were formed. Maybe there had been a loss of redundancy so that the old system was “breaking down” (whatever that means) but the language was a natural one, not a sort of conlang with exactly one means for every purpose.
the Uniformity Principle tells us that it’s very likely that there were other ways of identifying verb arguments (for instance) in concrete language where those combinations were formed.
The “uniformity principle” tells us no such thing. It says: “this is what I am assuming, and if you don’t agree you can buzz off.” Which is fine according to my lights, but let’s not pretend that it is more than that.
As to “other ways of identifying verb arguments (for instance) in concrete language where those combinations were formed”: I can imagine several non-linguistic ways, for instance pointing, punching in the face for not being clear etc. Also flouncing off in a huff – that’s what I do often.
From the Stanford Encyclopedia of Philosophy, The Problem of Induction[PDF]:
#
In the Treatise [of Human Nature], Hume says that
___if Reason determin’d us, it would proceed upon that principle that instances,
___of which we have had no experience, must resemble those, of which we
___have had experience, and that the course of nature continues always uniformly
___the same.
For convenience, we will refer to this claim of similarity or resemblance between observed and unobserved regularities as the “Uniformity Principle (UP)”. Sometimes it is also called the “Resemblance Principle”, or the “Principle of Uniformity of Nature”.
Hume then presents his famous argument to the conclusion that there can be no reasoning behind this principle. …
…
Bertrand Russell, for example, expressed the view that if Hume’s problem cannot be solved, “there is no intellectual difference between sanity and insanity”.
#
I find another another definition of uniformity that quotes from Chomsky:
#
In the absence of compelling evidence to the contrary, assume languages to be uniform, with variety restricted to easily detectable properties of utterances.
#
Whuuut? Bullshit. The principle of uniformity is just a special case of the parsimony principle. Induction has nothing to do with it.
What is “parsimony” good for, if not in reasoning about correlations and causality ? I assume you are referring to scientific arguments.
Anyway, the Russell quote was intended to be good for a laff, not for an outrage.
“These few examples, which could be multiplied endlessly, should suffice to demonstrate the point that [Chomskyan] Uniformity in general is not equivalent to Occam’s Razor. In syntactic theory, in order to simplify a theoretical account by making two superficially distinct objects the same, one must typically create some additional syntactic mechanism, syntactic structure, or syntactic representation in order to account for their differences. These additional mechanisms, structures, or representations are in general not independently motivated by empirical considerations, although it may be possible to reuse them for other analyses.” From Culicover & Jackendoff’s 2005 book _Simpler Syntax_.
These few examples, which could be multiplied endlessly, should suffice to demonstrate the point that [Chomskyan] Uniformity in general is not equivalent to Occam’s Razor. … These additional mechanisms, structures, or representations are in general not independently motivated by empirical considerations
By what then are they motivated ? Not only does that sound unlike Occam’s Razor, it sounds like the opposite – as cunningly implied by the formulation I have bolded. Entia sunt multiplicanda praeter necessitatem, cognitive convenience being the most important consideration.
@Stu.: As Jack Nicholson’s character didn’t quite say: “Forget it, Jake. It’s Chomskytown.”
I should watch that film again, I’ve forgotten everything about it.
It features nose slashing as a non-linguistic way of identifying a verb argument. Chinatown also expresses the moral view that it is unforgivable to force a fourteen-year-old girl to have sex with you, if she is your daughter.
cognitive convenience being the most important consideration
Zackly.
We Nominalists are free to regard all these fancy entia as mere* bookkeeping conventions.
Plato, Schmlato.
* I say “mere”, but good bookkeeping conventions are both unobvious and very useful. Inventing them is honorable and non-trivial work, not to be disparaged by wrongly characterising the invention process as “discovery.” Any fool can discover something.
I mean there is no Problem of Induction; induction does not reliably work, and this does not matter, because induction is not part of the scientific method. Parsimony is, and it can look similar, but it’s not the same. I’m surprised this escaped Russell – was Popper really the first to notice?
Chomsky paying lip service to parsimony and then ignoring or not understanding it is another story, and much, much less surprising.
because induction is not part of the scientific method.
What then is this “scientific method” to which you advert ? A modest technique for selecting and amassing true propositions, and nothing more ? Sounds rather Wiener Kreiß (they were a strident lot on the whole).
“Induct if you will, mais pas devant les domestiques“.
Not being a philosopher, the only Uniformitarian Principle (sue me) I know of is the one espoused by Ringe contra the idea of a universal European hydronymic language. I.e, if there are this many languages in Europe today, and even more in most other geographic areas of comparable size, it’s an error to assume that a single language covered all of Europe before PIE arrived. Likewise it’s an error to assume that you are seeing the whole picture if you reconstruct a pristine stage similar to a conlang for PIE, because no known human language today is like that.
Then apparently many people are in error. Can they be helped ? Must they want to be helped before they can be helped ?
I read that as a ‘universal European hydrodynamic language’ and my first thought was, that seems a bit specialized. Then my second thought was, why only Europe? Hydrodynamicists across the globe converse, of course, in broken English, the universal language of science.
The UP helps him who helps himself. It sufficeth unto me that I know how they should be helped. Endeavouring to make them want the help leadeth not my soul to calm waters, and thus I abstain.
A modest technique for selecting and amassing false propositions, approaching understanding of reality by a process of elimination. 🙂
@David Marjanović: Bacon felt that induction was the core of the scientific method—although he was very careful about what kinds of inductive arguments were licensed by observations. He did not quite express his method in the way Popper did centuries later, in terms of falsification of incorrect ideas, but the notion of induction as a process of ruling out what was inconsistent with real-world experience was definitely there—primarily negative, rather than positive, induction. Descartes was less focused on developing a specifically inductive method of gaining knowledge, which makes sense, since Descartes placed far more emphasis on mathematics in his analysis; the power of mathematical deduction as a tool in the inductive evaluation of scientific theories was something that Bacon apparently never appreciated. However, both of these “prophets of the scientific revolution” were reacting strongly against the centuries-long intellectual hegemony of the Scholastics’ almost purely deductive reasoning. So the notion of induction has become a key part of the language of the scientific method, useful for distinguishing modern investigations from the unscientific ideas that came before. Hume seems to have had no fundamental problem with this as a practical tool, but he pointed out that no kind of inductive argument, whether positive or negative, can be justified from logic alone—at least when it comes to predictions about how the world will behave in the future.
@Lars, well, I meant something else, that is “discernible meaningful elements in inflexion don’t mean that the previous stage was analytical”.
But yes, likely this entails presence of former morphemes which disappeared wihtout a trace.
Actually, Russian has shuffled its cases a lot, so there exists mechanisms that let people who need and use case to coin new meaningless elements (by analogy etc.). One can hypothesise that this is a normal behaviour for continuous users of case: they merge declension classes, they make this rhyme to that, they mispronounce, but they do not let any clitics in, period.
All right, it’s a hypothesis. If (if) it is true, there still a plenty of ways to let your clitics in. During a partial breakdown of the former system, or if the language retained overall structure that favours case, if suffixes replaced prefixes or new functions were coined etc.
A modest technique for selecting and amassing false propositions, approaching understanding of reality by a process of elimination.
Popperlapopp !
So it’s true that scientists investigate what is false, and will stay false reliably ? As Hume put it: that the course of nature continues always uniformly the same ?
It’s a bit of a shock to learn that research grants are given to discover false propositions, and that Nobel prizes are awarded for their discovery.
I think I’ll stay on the team of DE The Bookkeeper.
@stu
The false propositions are the ones not published (or published and overturned). Do you expect the miner of precious metals to display the surrounding dross with the metal or to find the metal without going through the dross?
…Oh. Texas German. 🙂
The false propositions are the ones not published
This is a major problem in medicine. Clinical trials that don’t support the effectiveness of a new drug are never published at all (through a combination of drug-company deliberate policy and lamentable journal editorial criteria) and it is difficult even to get funding for, let alone to publish, a trial meant to confirm or disconfirm an already published study (no matter how flawed.)
The actual ecosystem does not operate at all like Popper’s pretty fantasy. Avoidable deaths have resulted.
Science has means of generating frequently accurate propositions.
If we exclude this creative part from “science”, little will be left of it….
“The principle of uniformity is just a special case of the parsimony principle. Induction has nothing to do with it.”
I see how it can be connected to parsimony, but I also see induction.
We are speaking about extrapolation. I think it is inductive by definition. Presumably parsimony makes you prefer one extrapolation to another.
A person who knows the parsimony principle still can’t always reconstruct an equation from its graph or a segment of it (you see a Mandelbrot set, guess the formula and add a point), but she believes that among two equations that produce it one is better.
Does parsimony also tell that your graph can be extrapolated (that the ‘best’ extrapolation available is likely to be true)?
We of course, know situations when we don’t consider any extrapolation likely (you have one dot, it belongs to a graph of a simple function you studied in school. What’s the function? Oops) but for some reason we also can make some successful predictions about nature.
One if the problems David Eddyshaw mentions is known as the “file drawer effect.” Negative results get analyzed and, because they are hard to publish (or worse, because, say, a drug company may be strongly motivated not to release disappointing findings), they end up tucked away in a drawer somewhere, essentially unknown outside the lab where the experiment was conducted. In medicine in particular, where meta-analyses studying how well various drugs and treatments function in the aggregate, it is extremely important for all data collected about a particular treatment be available, not just a biased set of mostly positive results. Fortunately, medical research has largely recognized this, and many types of trials are preregistered and have their full data eventually made freely available, so that even if the results do not look interesting enough for a journal publication, they will not be held back from the community. (Preregistration also helps address with trolling through a data set looking for “statistically significant” correlations. The preregistered protocol describes what the data was originally collected to test. If something interesting turns up, that’s great, but it will be clearly identifiable as having been considered only after the fact, meaning it should be subject to greater statistical scrutiny.)
Pystynen’s blog discussed “anti-etymologies”, which cover both rejected etymological hypotheses and etymological dead-ends. I agree with him that those are useful and deserve to be published.
Fortunately that’s different in my field – though it wasn’t always, and it has little potential to save people’s lives except very indirectly.
Induction: The sun will rise tomorrow, because it always has.
Parsimony: The sun will most likely rise tomorrow, because gravity, inertia & stuff will make it so unless something very dramatic happens very soon even though all the obvious candidates would already be leaving evidence that we aren’t seeing.
…or unless the theory of general relativity is wrong in some pretty specific and very strange ways even though it has fared so extremely well so far; the ultimate incompatibility with quantum physics should not, as far as known unknowns go, be relevant here.
“[ ]: The sun will rise tomorrow, because it always has.
[ ]: The sun will most likely rise tomorrow, because gravity, inertia & stuff will make it so unless something very dramatic happens very soon even though all the obvious candidates would already be leaving evidence that we aren’t seeing.”
I’m not sure that the former is “induction” and the latter is “parsimony”. The latter uses a number of words introduced by modern science, both have some theory behind (“the sun” is already a theory) but the latter also has a theory of intertia and gravity.
I’m not sure what you mean, but maybe you mean parsimony when building models – not to predict anything, just to describe our observations. Induction then is extrapolating them to the unobserved.
If so, then uniformity pertains to induction.
Anyway, testing models is massively important for modern science. Some uniformity hypothesis, namely that we can identify a class of situations where our theory is supposed to work.
One can also make 1000 observations, and then share 500 of them with the researcher, to see if the model that the researcher would devise based on these 500 will also work for other observations. Then if the method for selecting this 1000 situations is reproducible, we expect the model to work for 1001s observation.
drasvi: ‘Next time someone says “voulez-vous coucher avec moi?” to me, I respond “there are no simple answers!”‘
DE: ‘Presumably, the more you know about the person asking, the more likely you are to answer like this.”
Oh, I misread it as “the more likely you are to answer positively” and wondered why is it at odds with both my present monogamous and teenage personalities.
I was asked this once. My (female) friend A told her (female) friend B that I know some 10 languages. In reality this was just the number of languages I dealt with (read something in), but before I clarified it, B asked if French is one of them and then said the only line in French she knows. My responce was boring.
(Years later we met again, found each other much more interesting, and even married once).
“Classical Arabic is a good example of a language with an elaborate case system which actually has an extremely low functional load.” – Not sure what’s elaborate, not the inventory of forms: just the three vowel phonemes. Some grammarian (Quṭrub < λυκάνθρωπος) even insisted those are added for euphony.
The more you know, the more you know that there are no simple answers.
Or to put it another way, the more you know, the less you know.
In the absence of compelling evidence to the contrary, assume languages to be uniform, with variety restricted to easily detectable properties of utterances.
That of course has nothing to do with uniformitarianism. It would be equivalent to Charles Lyell saying “Assume all rocks have the same chemical composition, with variety restricted to sizes, shapes, and maybe colors.”
It features nose slashing as a non-linguistic way of identifying a verb argument.
Not just verbs.
A modest technique for selecting and amassing false propositions, approaching understanding of reality by a process of elimination.
That said, the discovery of new white swans is not irrelevant to the truth of “All swans are white”. In particular, Wang Pei’s Non-Axiomatic Reasoning formalizes the notion that while induction and abduction are unsound, as the Schoolmen called it, they are still powerful tools: every white swan raises the likelihood that “All swans are white” is actually true.
every white swan raises the likelihood that “All swans are white” is actually true.
That “actually” ruins the probabilistic argument, which promotes gambling, not truth.
Every newly encountered white swan is an encouragement, at small risk of losing your shirt, to take a flutter on its not being the last white swan you will encounter.
NARS is not probabilistic in that sense, it is deterministic.
That said, the discovery of new white swans is not irrelevant to the truth of “All swans are white”.
A new white swan on the lake I walk past every Sunday has low relevance/only has relevance to “all swans on this lake are white on Sundays”.
That’s why science tries to reproduce results on different continents, different times of year, different … anything that might or might not be relevant.
As opposed to for example assuming that a measurement of the earth’s circumference in BC Alexandria (by a now-lost method) is the once-for-all size of the earth. (Why set sail Westwards on such non-reproduced evidence, when it would be easy to repeat the experiment?)
NARS is not probabilistic in that sense, it is deterministic
I know only one sense of not probabilistic, namely deterministic.
Of course, if words are to mean just what we choose them to mean, neither more nor less, then I can probably not determine what you’re on about.
they often claim that interference from a second language which has similar grammatical categories but a different way of marking them morphologically may lead to morphological simplification
Precisely how Trudgill explains the zero 3sg ending in Norfolk dialect, with help from Those Whom Nobody Expects.
primrose path to pseudoscientific perdition.
Best pronounced by restoring the missing Greek π.
pseudo-pscientific
I saw my first black swan (πswan?) today. (Parque del Buen Retiro). So now I can prove anything? There are some pesky conjectures I’ve got my eye on…
At least I think it was a swan. The side facing me was very black, that much I’m sure of. And it looked a lot like those Australian ones on Wikipedia.
those Australian ones on Wikipedia
They’re also common in New Zealand. I saw a pair on my local duck pond last week. They’re aggressive blighters. Must’ve come here from Australia.
As well as the black feathers, they have noticeably red beak and legs.
The one I saw had black feet, in fact. And so do the white ones in Denmark, but for some reason I think of them as orange like the break. Aggressive buggers too, especially the males.