Veronique Greenwood writes for the Atlantic on an ever-interesting topic, A Language’s Popularity Could Influence Its Grammar and Vocabulary:
It’s a peculiar observation that the more people speak a language, the simpler its grammar tends to be. English and Mandarin, for instance, have notably straightforward structures. On the other hand, languages spoken in just a single mountain valley or village can have gorgeously intricate grammars, full of gender and cases and declensions. They also tend to have rather small vocabularies. Meanwhile, the vocabularies of widely spoken languages are enormous.
What is going on here? What connection might there be between how many people speak a language and what it is like?
There are many things that could be at play, from the level of historical trends all the way down to how parents speak to their children. It also isn’t exactly clear which comes first: lots of speakers, or simple grammar? Still, the researchers behind one recent paper wondered whether the fact that grammar is relatively hard to learn and new words relatively easy might be enough to explain this trend, at least in broad strokes. They built a mathematical model in which individuals in small and large social networks have conversations and occasionally learn new words or ways of saying something from each other. What the team found was that even in this simple, stick-figure version of the world, the same patterns emerge. The results suggest that these general rules, along with the number of speakers, can influence how a language grows and changes.
Read further for details and suggested explanations; if you want to dig deeper, here‘s the actual paper, “Simpler grammar, larger vocabulary: How population size affects language” by Florencia Reali, Nick Chater, and Morten H. Christiansen (Proceedings of the Royal Society B 285, issue 1871). Thanks, Bonnie and Martin!
If I didn’t know better, I’d suspect you of trolling, Hat. English and Mandarin have “notably straightforward structures” now? Geoff Pullum and Yuen Ren Chao would be relieved. And they have an accepted metric for overall grammatical complexity of a language since I last heard?
I’ll resist the urge to cite twenty counterexamples (in both directions) off the top of my head. We all could.
Yet another GIGO computer simulation.
Yeah, I’m afraid it’s problematic. But interesting!
To be a bit less intemperate (but where’s the fun in that?) it does seem to be the case that morphologically exuberant languages (which is presumably what they have in mind) tend to be found more often in smaller language communities. But the samples in experiments like this are ludicrously unrepresentative of world language diversity, and moreover by historical accident most “big” languages belong to only a handful of genetic families, potentially skewing the results enormously.
As far as vocabulary goes, I don’t trust any assessment of total vocabulary size in the language of a small community unless it’s made over more than twenty years by more than twenty native speakers well versed in all traditional genres.
[Actually, I think I’m being too kind: unless I’ve misread the paper (sadly possible) there aren’t any actual data. It’s a glorified thought experiment.]
Correct: it’s a computer simulation.
Who can forget the picturesque isolated mountain valley where Swahili speakers weave their intricate phrases hidden from the outside world.
Yes, I’m assuming by ‘grammar’ something like ‘morphology’ is meant.
A semi-counterexample is the Chamic language family, which looks like (modulo sound change) a large and widely spoken language morphologically and syntactically despite having only 4.5 million speakers (of which 3.5 million speak Acehnese). The reason is that the Champa Empire used it as its lingua franca, and most of the people who spoke it were descended from L2 speakers (as indeed is the case for Bengali and Swahili).
Linguistic innovations that are relatively easy to learn …
This seems to come back to the Hofstadter piece on machines treating language as just patterns of sounds/letters, with no internal model of what anything means/what language is used for.
How do “linguistic innovations” appear? Perhaps somebody brings a sack of new-fangled potatoes that a community thereby needs a word for. Why use (or mangle) a word that comes with the item vs inventing a compound from words already in the language (pomme-de-terre)?
If the so-called research is actually focussing more on morphology, why does anybody need to invent new morphology? In what way are they modelling language contact: a group of speakers arriving?
Yes indeed, English is gloriously straightforward, as long as you never use verbs.
By “complex grammar” they appear to mean “synthetic morphology”. Hm.
But most Romance languages are fairly synthetic, though not at the extreme, right? And aren’t Hindi and Urdu for another widely-spoken example?
I guess I should read the thing. Probably it just claims an effect in this direction without caring if it’s tiny, eh.
small linguistic communities favour linguistic innovations that are hard to learn (such as, we suggest, structural changes in the language)
I wonder if this statement is actually based on research or is simply the hypothesized counterpart of “large communities favour innovations which are easy to learn”.
In fact, small, geographically isolated communities (not sub-communities such as are found in large cities) tend to preserve linguistic archaisms which are barely surviving or have been regularized in larger communities which attract new speakers. That such small, culturally conservative communities would create structural innovations, and make those innovations hard to learn, does not make sense. I agree with AntC above: why does anybody need to invent new morphology? (if they continue to speak the same language and do not add new adult speakers).
What is “a large linguistic community” anyhow? At least until a few hundred years ago, most communities were small. Some might hardly ever interact with neighboring communities, even if their speech varieties were nearly identical.
This model is fine as far as it goes, but the phenomena it purports to explain may be imaginary.
(I should qualify this. Reali et al. compare populations between 20 and 500 people, which is a reasonable range.)
To be a bit less intemperate (but where’s the fun in that?)
The opposite of ‘intemperate’ is…
Stodgy
“To be a bit less intemperate (but where’s the fun in that?) it does seem to be the case that morphologically exuberant languages (which is presumably what they have in mind) tend to be found more often in smaller language communities.”
David, I’ve noticed something like this but not so much about big, intricate morphologies as just tangled masses of exceptions and lots of lexicalized phrases – I’m thinking of Irish as compared to English. German is similar in this way too – and it has more to do with people learning these languages as L1s – kids will do ANYTHING to fit in – as opposed to L2 learners – there is some shit adults just won’t put up with.
Except….Navajo added a huge number of L2 speakers during that century when the Comanche moved in and demolished pueblos on the Pecos and created waves of refugees. The Navajos took them in and these people learned Navajo and seem not to have had much effect on the morphology.
This paper by Peter Trudgill is, IIRC, somewhat related:
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.923.6533&rep=rep1&type=pdf
@Jim:
Good point. It’s a whole third dimension of language complexity apart from vocabulary and morphology, and it’s even more difficult to think of how it could be measured in any objective way.
I happened to be reading John Haiman’s account of Cambodian, certainly a big language and one which fits the supposed rule inasmuch as it’s thoroughly inflection-free (there’s a good bit of derivational morphology but not much of that is productive, even.) But he talks about it being a classic case of a “Desesperanto”, “a language of which one can read a page and understand every word individually, and have no inkling of what the page was all about”, because so many phrases aren’t analysable as just the sum of their parts.
I think Latin is a bit like that in relation to Classical Greek; much smaller total vocabulary, but concepts get expressed via unobvious metonymy and opaque phrases whose meaning can’t be guessed at just by looking up each word in a vocabulary.
The Navajos took them [Comanche refugees] in and these people learned Navajo and seem not to have had much effect on the morphology.
The Comanche refugees presumably included whole families, whose children would learn Navajo by playing with Navajo children and serve as interpreters between their parents and the Navajo adults, who would still have been the dominant ones while the Comanche were in a socially subservient position until they thoroughly amalgamated with the Navajo (I don’t know what the actual history is). Even if the two populations had been on a really equal footing, Navajo, a Dene language, and Comanche, a Uto-Aztecan one, are extremely different and it would be difficult for one of the languages to affect the morphology of the other one.
“Desesperanto”: “a language of which one can read a page and understand every word individually, and have no inkling of what the page was all about”
A lot of morphology-lite languages look like this. It’s almost inversely correlated with morphological complexity, I think.
Come to think of it, Classical Chinese must surely be the ne plus ultra of Desesperantoj.
I knew the expression “a language of which one can read a page and understand every word individually, and have no inkling of what the page was all about” was reminding me of something.
From section 6 of Why Chinese is So Damn Hard. Also contains the memorable phrase: “I think it’s about a phoenix or something.”
Slightly OT, but…
>Whereas modern Mandarin is merely perversely hard
Mandarin being hard seems to boil down to:
1. Its orthography is hard. No comment needed.
2. It’s very different from English, the lexicon is different, tones are weird…
3. It occasionally sticks in classical Chinese, which is hard.
But if 1. you focus on spoken Mandarin 2. you’re a Martian 3. you accept classical Chinese references as something you just have to memorize, is contemporary Mandarin really that hard?
I feel like one thing it has going for it is a high ratio of native and/or reused morphemes. Chinese people who study English just have to accept the fact that a lot of English words consist of French/Latin/Greek morphemes. Sure, the way those morphemes are part of the language is not _completely_ unsystematic, but it still seems like a big burden. I would feel much more discouraged in studying Mandarin if I was running into de-facto-cranberry superstrate morphemes all the time.
What is “a large linguistic community” anyhow?
It seems to me that “large”/“small” is a bit jargon-y here. The key is their assumption that the number of people each speaker knows “grows superlinearly with . . . population size.” This seems reasonable and they have evidence for it, but if I understand correctly the mechanism is more driven by that connectedness than by largeness as such. Speak a lot to a few people, be exposed to fewer innovations but with more reinforcement per innovation. Speak a little to a lot of people, be exposed to more innovations but less reinforcement.
But if 1. you focus on spoken Mandarin 2. you’re a Martian 3. you accept classical Chinese references as something you just have to memorize, is contemporary Mandarin really that hard?
I think you’re right about that, but, to be fair, the quoted essay (which is indeed delightful!) does make this clear at the top. “I mean hard for me — and, of course, for the many other Westerners who have spent years of their lives bashing their heads against the Great Wall of Chinese.” The author lists nine reasons why Chinese is hard; five (!) come down to its the writing system, two are Westerner-specific linguistic complaints (lack of cognates, tonality), and two are Westerner-specific cultural complaints (Classical Chinese, lack of shared cultural references). Seems about right.
Probably not. What little I actually know and have noticed about the grammar = syntax is that some of it is very different from Standard Average European, but straightforward once you’ve managed to wrap your mind around it.
Classical Chinese isn’t that impossible either, once you figure out that it’s several dozens of different languages. “Chinese of 7-9 century highbrow prose”, “medical Chinese” etc etc are individually humanly possible; there is just very little transferable skill between styles. Once you learn a dozen of them, you claim that you know Chinese.
@ marie-lucie:
The Comanche refugees presumably included whole families … Navajo, a Dene language, and Comanche, a Uto-Aztecan one, are extremely different …
A relatively minor point, but I took Jim Doyle’s comment (hopefully he can correct me if I’m wrong) as saying that the Comanches attacked Pecos pueblos, and the inhabitants of the latter — presumably not Comanches themselves, and thus not speakers of Commanche or Navajo — became the refugees who then moved in with the Navajo.
They seem to have spoken Tanoan languages, so perhaps extremely remotely related to Comanche, but not related to Navajo at all (not even if you’re Greenberg.)
Insofar as the list of Navajo clans (which are matrilineal) is reliable, there are clans claiming descent from the Paiutes, the Chiricahua Apache, the Mohave, the Mexicans, the Mescalero Apache, the Zia Pueblo, the Jemez Pueblo, the Pima, the Zuni Pueblo, the Utes, and the San Felipe Pueblo. African Americans do not constitute a clan, but Navajo with African American fathers will say they are “born for the Black People”. There are Navajo terms for whites and Asians, but they don’t function in the same way.
Someone needs to tell those 330 million Hindi/Urdu speakers that they all need to be living in a single mountain valley because of their language’s gorgeously intricate grammar, full of gender and cases and declensions. While we’re at it, let’s tell the Russians (150 million speakers) and Germans (100 million speakers) as well.
Written Standard German has more cases and tenses than the spoken language (never mind the related languages, standardized or not), and I suspect the same is true for the other examples.
That’s not true about the tenses north of the White Sausage Equator. It pretty much is about the cases, though, and I doubt the “present subjunctive” (Konjunktiv I) is used in halfway spontaneous speaking anywhere.
“who would still have been the dominant ones while the Comanche were in a socially subservient position until they thoroughly amalgamated with the Navajo (I don’t know what the actual history is). ”
M-L, I was unclear – these were refugees from the Comanche onslaught. They probably spoke Tanoan or maybe Keresan languages originally.
Thanks, Peter, for making sense of my muddle.
Eddy, that’s right.
“What little I actually know and have noticed about the grammar = syntax is that some of it is very different from Standard Average European, but straightforward once you’ve managed to wrap your mind around it.”
David, that’s true in general. Lushootseed grammar is a model of pellucid clarity once you just stop expecting to see the wrong things in it.
John, that’s a really interesting ethnographic point. It mirrors the situation in modern American society in a way.
There is also a clan called Many Comanche Warriors, but I assume that meant ‘defeated many Comanche warriors’ rather than claiming descent from them.
David: I should have said tenses/moods.
Peter Erwin: … the Comanches attacked Pecos pueblos, and the inhabitants of the latter — presumably not Comanches themselves, and thus not speakers of Commanche or Navajo — became the refugees who then moved in with the Navajo.
I read and wrote too fast! You are absolutely right, I was absolutely wrong about who was who. But my argument is still valid as to who was in what position socially and linguistically..
Apologies and thanks to you too, Jim Doyle.
David E: They seem to have spoken Tanoan languages, so perhaps extremely remotely related to Comanche, but not related to Navajo at all (not even if you’re Greenberg.)
Sapir had put together “Aztec-Tanoan”, perhaps because of a few words which are probably borrowings. Very few people still consider this “phylum” credible at all.
JC: Insofar as the list of Navajo clans (which are matrilineal) is reliable, there are clans claiming descent from the Paiutes, the Chiricahua Apache, the Mohave, the Mexicans, the Mescalero Apache, the Zia Pueblo, the Jemez Pueblo, the Pima, the Zuni Pueblo, the Utes, and the San Felipe Pueblo.
The Navajo, speaking an Athapaskan language (therefore genetically linked with languages spoken mostly in Alaska and Northern Canada, and perhaps also in Siberia!) are supposed to have migrated South (along with the Apache, who are also Athapaskans), and then must have amalgamated with local peoples speaking a variety of languages from several different families, probably causing a fair amount of borrowings.
> David, that’s true in general.
There’s a difference between something that is very different from what you’re used to, but which is inherently simple once you overcome the differences, and something complex with lots of exceptions and special cases, which might seem familiar at first sight. Japanese is consistently left-branching, which might seem alien to SAE speakers. Mandarin is mostly left-branching with the VO order being the main exception. English branches according to the wind (I’m kidding of course, but the rules are more complicated).
Drifting off topic again, but I recently realized English almost has serial verbs just like Mandarin when I told my son: “Go help Mommy find a box”.
I, too, question the premise of this article. English and Mandarin aren’t the only widely spoken languages in the world and it doesn’t seem obvious to me that English’s recent rise to the status of international lingua franca caused British English to become significantly more simple than it was in, say, the 18th century.
It is perhaps more arguable that English and Mandarin have both experienced standardization processes, which would have reduced esoteric structures, but as far as I know, modern English already had its present structural properties in the 18th century, before what we might expect to be significant back influence from colonial English.
@John Cowan: Colloquial Russian has the same number of cases and tenses as written standard Russian (the number of tenses is small anyway, there are only three). That is due to two developments – the historical development of the written standard went from it actually being a different language (Church Slavic) towards getting more similar with the spoken language, and then in the 20th century the standard language crowding out most of the regional and local variety. Differences between written standard and colloquial are mostly in the area of lexicon, syntax, and in details of morphology and usage of categories, not in the number of items per category.
Eidolon: “[But] as far as I know, modern English already had its present structural properties in the 18th century, before what we might expect to be significant back influence from colonial English.”
Indeed. English lost almost all the features of Old English – “gorgeously intricate [grammar], full of gender and cases and declensions” – while it was still a relatively minor Germanic language spoken on the periphery of Europe long, long before it attained any global reach.
dianichi,
“English branches according to the wind (I’m kidding of course, but the rules are more complicated).”
…which has to do with how much cauliflower it’s eaten recently… But you’re right and it’s a feature of SAE languages in general, probably a residual effect of so much case marking and gender and verb agreement that lent s much flexibility to the earlier stages of these languages.
“Drifting off topic again, but I recently realized English almost has serial verbs just like Mandarin when I told my son: “Go help Mommy find a box”.
English shares the pivot construction with Mandarin – “make him go” – that other SAE languages do not have.
And if you look at the second formatives in phrasal verbs as if they are verb-like, you find that Mandarin uses verbs. So semantically at least English phrasal verbs can be seen as serial verbs.
Perhaps because it was being learned by large numbers of Vikings at the time.
(Edit: not original to me; apparently original to John McWhorter.)
“Perhaps because it was being learned by large numbers of Vikings at the time.”
As well as a far number of Britons, after all.
> But you’re right and it’s [complicated left/right branching rules] a feature of SAE languages in general
I think I read somewhere that German is actually pretty consistently left-branching, except there are some very big exceptions, namely V2 in matrix clauses and post-positional relative clauses. Participal adjuncts (as in “die in meiner Hemat gleich nach Ostern geschorenen Schafe”), which I believe are more common in German than the average SAE language, are also left-branching.
> probably a residual effect of so much case marking and gender and verb agreement that lent s much flexibility to the earlier stages of these languages.
If my understanding is correct, PIE was mostly left-branching (and SOV), but with a certain degree of free word order. As agreement became less prominent in many descendant languages, right-branching became more prominent to make up for this (both to use position as a POS marker, and to avoid dependents drifting too far away from their heads).
As well as a far number of Britons, after all.
Huh, I’m not sure that had ever occurred to me.
Perhaps because it was being learned by large numbers of Vikings at the time.
The Norwegian linguist Jan Terje Faarlund and a colleague think that the ancestor of Modern English is not “Old English” (also known as Anglo-Saxon) but a Scandinavian variety brought by the Viking settlers of England. The languages would probably have been largely mutually intelligible at the time.
Extraordinary claims require extraordinary evidence.
Thanks for the link JC.
That was quite interesting to reread.
“Extraordinary claims require extraordinary evidence.”
At least no one is suggesting that Germanic is “undemonstrated” because of all these areal effects, unlike Penutian for example.
@Jim Doyle:
> English shares the pivot construction with Mandarin – “make him go” – that other SAE languages do not have.
German has “ihn gehen lassen”, French has “le faire partir”, Spanish has “hacerlo ir”… How are they different?
I was playing fast and loose with nomenclature, I should have said serial and pivotal verbs. In Danish, “Go help Mommy find a box” would be “Gå over og hjælp mor med at finde en kasse”, literally “Go over and help Mommy with to(inf.marker) find a box” with explicit “serializing” and “pivoting” structures.
In French “aider à trouver”, I guess one can debate whether à is a preposition or a kind of infinitive marker, but either way, overall English strikes me as having less of these verb-connecting structures than other European languages.
Strikes me as a nominalization. Indeed, beim Schachtelsuchen (yay noun incorporation!) would be an acceptable German translation.
Jim Doyle: At least no one is suggesting that Germanic is “undemonstrated” because of all these areal effects, unlike Penutian for example.
Indeed, but Germanic is a small, obvious family, while “Penutian” is a still controversial group of 15-odd such families, so in terms of scale it should be compared to Indo-European as a whole rather than to any of the component IE families. Comparative-historical studies in Penutian cannot really be evaluated against those in Indo-European, for a variety of reasons. But many features which occur in several Penutian families are attributed to “areal effects” rather than to common origin.
dainichi: French le faire partir, aider à trouver
These snippets of sentences sound odd without a context.
For the first example, faire + infinitive ‘to cause s. to …’ seems to me to be an instance of a pivot verb, followed by a verb in the infinitive carrying the lexical meaning. It occurs in two structures, one where the lexical verb is intransitive and the other where it is intransitive. The verb faire ‘to do, make’ in these structures has a causal meaning which it does not have when it is a main, lexical verb. The only other verb which occurs in a similar structure is laisser ‘to let (s. do, happen, etc)’.
For the second example, as in Va aider Maman à trouver une boîte, there is no specific structure aider (s.) à + infinitive: aider ‘to help’ can be transitive or intransitive, always has the same meaning and (if present) the à + infinitive sequence is a complement of this verb, occurring after its direct object. The preposition à is certainly not an ‘infinitive marker’, since French infinitives (like Spanish, Italian, etc ones) are marked by suffixes.
But many features which occur in several Penutian families are attributed to “areal effects” rather than to common origin.
marie-lucie, could you elaborate on that? Which features, and where would that contact have occurred?
nominalization: Not really, or it depends on your definition I guess. at finde with complements works as a free NP, unlike the unadorned infiniitve which only occurs as the complement of certain verbs, mostly modals, but it’s not a noun.
If you want a full-fledged noun that can take determiners and inflect for definiteness, your best bet is derivation with -eri where you can compound in the object: Hjælp mor med hendes kassefinderi.
There is also the old verbal noun in -en, but for obscure reasons it only really works with verbs for vigorous activities like running, screaming and shouting, at least in my idolect. Lad os undgå løben på gangene = ‘Let us avoid running in the halls.’
I didn’t really catch what the distinction between serial and pivot verbs was, but I’m not sure Da lade is either anyway, it feels more like a kind of modal.
@David
> Strikes me as a nominalization
As Lars says, it depends on your definition. Danish at-infinities and at-subclauses can be more NP-like than their equivalents in other languages, since they can be ruled by prepositions. This means that verbs can be more consistent in how to take their arguments, e.g. you help somebody with a problem, and you help somebody “with to” find a box.
@m-l
> These snippets of sentences sound odd without a context.
Sure, I was trying to translate Jim Doyle’s “make him go”, which I chose to read as an infinitive, not an imperative.
> The preposition à is certainly not an ‘infinitive marker’, since French infinitives (like Spanish, Italian, etc ones) are marked by suffixes.
Danish infinitives have suffixes, namely “-e”, but “at” is still an infinitive-marker. “At” is a cognate of French “à”, by the way.
I don’t know which syntactical tests are used to determine whether a particle is an infinitive-ruling preposition or an infinitive marker. But there are only a few prepositions (à, de) which can appear before infinitives in French, and they usually just provide the function of allowing the infinitive to become a verbal adjunct.
Oh, just noticed David’s Schachtelsuchen which of course would be the same inherited derivation as Da ?kassefinden and E box finding. And against Danichi, you can actually keep the argument structure in English: help sby with finding a box.
(But as I said, the -en verbal noun has very restricted use in modern Danish, compensated by the freer use of the at-marked infinitive. There may be causation there, but which way I would not even try to guess).
I seem to remember that Pullum proposed that “to” was an auxiliary verb. I can’t find a reference, though.
Classical Chinese isn’t that impossible either, once you figure out that it’s several dozens of different languages.
On what level is that? Different syntax? Different vocabulary? What is not transferrable between them?