Difficult languages and easy languages, part 3
« previous post | next post »
There may well be a dogma out there stating that all languages are equally complex, but I don't believe it, especially not if it has to be "drummed" into our minds. I have learned many languages. Some of them are exceedingly hard (because of their complexity) and some of them are relatively easy (because they are comparatively simple). I have often said that Mandarin is the easiest language I ever learned to speak, but the hardest to read and write in characters (though very easy in Romanization). And remember these posts:
"Difficult languages and easy languages" (3/4/17)
"Difficult languages and easy languages, part 2" (5/28/19)
Languages are difficult /easy // complex / simple in different ways. It is reductionist to say that they are equally difficult /easy // complex / simple.
Selected readings
- "Why Chinese Is So Damn Hard" (8/91)
- (简体字：为什么中文这么TM难？)
- (繁體字：為什麼中文這麼TM難？)
- "Is Mandarin easy to learn after all?" (5/29/19)
- "Why Literary Sinitic is so darn hard" (5/30/10)
- “Which is harder: Western classical languages or Chinese? ” (3/6/16)
- "Which is worse?" (1/21/16)
- "Difficult tongues" (5/14/21)
- "The implications of Chinese for AI development" (10/25/21)
- "The implications of Chinese for AI development, part 2" (10/27/21)
Daphne Preston-Kendal said,
October 29, 2021 @ 9:56 am
For the opposite view of the difficulty of learning Chinese, from someone who has great motivation as it’s his family’s native language, see Raymond Chen: https://devblogs.microsoft.com/oldnewthing/20141106-01/?p=43673
For what it’s worth, though I’ve never tried to learn Chinese or another heavily analytic language, I agree with Chen that inflections are a godsend for catching one’s own mistakes. Especially when a foreign script is involved: when learning Sanskrit I noticed while translating one sentence that what I had transcribed from Devanagari into Latin read ‘The man are going …’. So I double checked, and yes, I had misread a vowel. (Probably नराः as नरः, for those of you following along at home.)
So while there may be no objective measure of how difficult a language is, I don’t think anyone can deny that it can vary subjectively from person to person. But it probably depends on personal taste at least as much as similarity to whatever languages someone already knows.
Victor Mair said,
October 29, 2021 @ 10:29 am
@Daphne Preston-Kendal:
Thanks very much for your perceptive comments, also for calling our attention to Raymond Chen's excellent blog, which probes the complexity that is hidden within the simplicity of Mandarin.
This also raises another key issue: why children of parents who speak L1 often give up learning L1 and just go with the L2 of the environment in which they find themselves living and growing up. This is a psychological dimension that is seldom discussed. Namely, subconsciously or even consciously, L1 speakers often rebel against and essentially refuse to learn the language of their parents when the latter force / coerce them to do so. I have observed over and over again that children who are pressured by their parents to learn L1 — especially if that involves belittling them for not doing so — often harbor deep resentment against the assumption that — although living in an L2 society — by accident of birth, it is demanded of them that they should be fluent in L1.
Daphne Preston-Kendal said,
October 29, 2021 @ 11:05 am
Here in Germany, I hear similar sentiments from English-speaking parents of children growing up here. E.g. the children answer in German even if their parents start talking to them in English. There can be extra reasons to resent having native English skills, as at least some children nonetheless have to sit through their English-as-a-foreign-language classes at school for bureaucratic or organisational reasons. (‘The curriculum requires English!’ or ‘We’d have to find special activities for them in the timetable while everyone else in their class learns English!’) That must be extremely boring and frustrating.
However, in the case of English, the resentment tends to last only until their teenage years when their peers’ English skills reach the level where they can start to enjoy English-language popular culture together. At that point, they find their special English abilities cool and speak it more (including with their parents) and become more comfortable with code-switching, for example. (Some adults I know who grew up here with English-speaking parents code-switch like crazy, within practically every sentence, even when speaking to native German speakers — though of course my presence in such conversations, as a native English speaker, could influence that as well.)
Unfortunately, languages without the cultural cachet of English likely suffer more. (Germany’s poor record on preserving its own minority/regional languages, such as North Frisian, may well be caused by similar dynamics.)
DJL said,
October 29, 2021 @ 11:31 am
But isn't the 'all languages are equally complex' dogma a point about linguistic theorising – ie, about the theoretical tools needed to describe/explain linguistic structure?
I don't think I have ever encountered the 'dogma' being spelled out in terms of 'ease of learning', let alone in the context of second language learners, which is what you seem to be getting at.
J.W. Brewer said,
October 29, 2021 @ 12:52 pm
While I do suspect that the pseudo-spiritual axiom that "all languages are equally complex" is too pat to be completely true, I think a more significant issue is that any given language may be comparatively easy to learn for someone with such-and-such L1 yet quite difficult to learn for someone with such-and-such else L1 which is more radically dissimilar (in multiple dimensions) from the proposed L2. So I don't know where the neutral/baseline starting point is from which you could judge the "objective" difficulty of learning a given proposed L2 without knowing the L1 of the would-be learner.
Scott Mauldin said,
October 29, 2021 @ 1:43 pm
To respond to J.W. Brewer and defend Professor Mair's position, while I think what you are saying is true, J.W., the extent of that "ease by similarity" can still vary enormously between groups of these similar languages. Like there could be four Amazonian tribal languages that are very similar and learners of each one can learn the others very easily, but all could be nigh impenetrable for most language speakers in the rest of the world. I think you'd need to look at the "average difficulty for a speaker of any world language". A language like Mandarin (which I agree has a very low barrier to entry) could be very low-difficulty on average, even if it's harder for Arabic speakers than English speakers, for example.
LN said,
October 29, 2021 @ 2:45 pm
I agree that the difficulty of learning a language depends on many factors. For L1 English speakers, there is a language difficulty scale from the FSI https://www.state.gov/foreign-language-training/
Since Chinese and Vietnamese belong to a Sprachbund, I think it is interesting to compare these two languages. Chinese is classified as a Category VI ("super difficult") language, while Vietnamese is a Category III language. Because of the difficulty of the characters, the FSI assumes that twice as many hours are required to learn Chinese compared to Vietnamese.
As for the statement "all languages are equally complex", I have seen this more often in the context of information theory. If two speakers of different languages want to convey the same information to a listener, then they can do so in the same time. If this were not possible, the language would adapt (structure more "difficult" / less "difficult").
S.H said,
October 29, 2021 @ 3:48 pm
"This also raises another key issue: why children of parents who speak L1 often give up learning L1 and just go with the L2 of the environment in which they find themselves living and growing up. This is a psychological dimension that is seldom discussed."
The late psychologist Judith Rich Harris famously proposed and argued (which was controversial in psychology at the time) that peer groups have a stronger impact on kids growing up than parents. One of the lines of evidence she used was indeed was psycholinguistic, that immigrants' kids' accents and language, among other things, resemble their peers not their parents.
Kids get cues growing up from peers about how to act in the wider world as they come of age. Kids might learn stuff from their parents in their home environment different from their outside environment, but since eventually they will have to leave home and function in the wider society, it makes sense to be cued in as what's "normal" in the wider society rather than just "what I do at home, with mom and dad, but which I observe others not really doing".
I suspect in many cases when they and their families are the only speakers of a language in town, immigrant kids may not have a large enough peer group (outside the home) as speech community to sustain its use.
"I have observed over and over again that children who are pressured by their parents to learn L1 — especially if that involves belittling them for not doing so — often harbor deep resentment against the assumption that — although living in an L2 society — by accident of birth, it is demanded of them that they should be fluent in L1."
This isn't particular to language too, though, isn't it? Parents doing things differently from the wider society (e.g. parents who are trying to instill conservative culture in kids while the wider society is more liberal, or parents raising their kids vegetarian in a society where vegetarianism is seen as weird, or practicing a religion different from the mainstream) can receive the same rebellion by their kids and resentment that "why is it only my family that has to follow (whatever the rules are). "It takes a village to raise a child" is kind of cliche'd but it is quite relevant.
Jerry Packard said,
October 29, 2021 @ 3:52 pm
Perhaps a reasonable measure of how difficult it would be to learn any given language would be how much a child is able to learn of the language by a fixed age, say, 36 months. The proof that some languages are more difficult than others then would be that some languages — the 'more difficult' ones — would be found to be less completely learned by the end of the time period.
cliff arroyo said,
October 29, 2021 @ 4:09 pm
"why children of parents who speak L1 often give up learning L1 and just go with the L2"
All else being equal… children want to speak the language of other children. The language of the playground usually trumps mom and dad's language (when the two are different).
"children who are pressured by their parents to learn L1 — especially if that involves belittling them for not doing so"
I can't imagine that being belittled by parents would or could create any enthusiasm for what is, for children, a foreign language. it sounds like rotten parenting, to be honest.
AntC said,
October 29, 2021 @ 4:14 pm
"average difficulty for a speaker of any world language"
I'm half-remembering a quote from Anthony Burgess, to the effect you can become proficient very quickly in Malayan for day-to-day purposes, but it'll take a lifetime to master it.
I have often said that Mandarin is the easiest language I ever learned to speak
So we appear to be lacking an explanation why AI is struggling in the case of Mandarin — that is, if we are comparing apples with apples, as I said on the earlier thread.
KWillets said,
October 29, 2021 @ 4:38 pm
My wife and I worked for a long time to make sure that our kids wouldn't just listen to Korean and respond in English. Now, the effort has finally paid off — she speaks to the kids in Korean, and I respond in English.
But seriously, the L1/L2 thing starts around pre-school when kids realize they can talk to other kids in L2, and L1 stays at the pre-school level. We were lucky to find a bilingual program where they could continue to develop functional vocabulary. They also visit family and keep up relationships with them.
J.W. Brewer said,
October 29, 2021 @ 4:40 pm
@AntC: because the AI isn't working with audio of spoken Mandarin, it's working with written Mandarin, which Prof. Mair did not claim was easy to learn.
AntC said,
October 29, 2021 @ 5:41 pm
dogma … that all languages are equally complex, but I don't believe it, especially not if it has to be "drummed" into our minds.
The 'drumming-in' was on a very introductory Linguistics course (in the 1970's). I guess the instructors felt they should counter the cultural imperialism that Classical and European languages are somehow more 'sophisticated'; whereas languages without a word for 'wheel' or 'steamship' or trans-global navigation thereof couldn't be so "complex".
Is there an echo in the Everett/Chomsky debate about Pirahã? That a language not exhibiting recursion/embedding (_if_ it doesn't) cannot be so "complex".
Alexander Pruss said,
October 29, 2021 @ 8:37 pm
Should one assume that the alleged dogma is restricted to natural languages? For once we include constructed languages, we have languages that are designed for ease of learning and presumably some of them succeed at this, and some that are really hard (Googling turns up Ithkuil which allegedly has no fluent speakers, not even the creator).
And we can certainly see that there is a variety of ways of a language being easy or hard. Thus, the syntax of First Order Logic is really easy, and the vocabulary of a particular dialect may be super easy, but it is really hard for a student to translate something like "Max fed all of his pets before Claire fed any of hers" into FOL (I only gave two or three grad students full points on this exercise in over a decade of teaching, and I am not confident that those translations correctly covered all the cases where Max and/or Claire have infinitely many pets).
John Swindle said,
October 29, 2021 @ 9:00 pm
Doesn't AI do fairly well at converting Mandarin Chinese text to pinyin? Shouldn't that make Mandarin an easy language for AI to conjure with? If conversion the other way around is less reliable, make it talk out loud when it has to deal with humans.
James Wimberley said,
October 30, 2021 @ 7:09 am
Assume for the sake of argument that spoken languages are all equally difficult. The proposition does not extend to the written versions, since languages vary in the weirdness of their orthography. W-Spanish is clearly relatively "easier" than W-French, with its homophones creating a dual grammar, and W-English with its spelling chaos.
Jerry Packard said,
October 30, 2021 @ 9:26 am
There is indeed an echo in the Everett/Chomsky debate about Pirahã.
I'm not sure what all the hoopla is about Mandarin being difficult. Whether in pinyin text or audio transcripts, Mandarin is one of the easiest languages, at least for translation. For character texts – where the issue becomes automatic word parsing – it still gets translated with a fairly high accuracy rate.
Philip Taylor said,
October 30, 2021 @ 10:02 am
Jerry — "I'm not sure what all the hoopla is […]" — <Am.E> "hoopla" = <Br.E> "hoo-ha" ("a commotion, a rumpus, a row") ?
Jerry Packard said,
October 30, 2021 @ 10:13 am
@Philip Taylor
= hoo-ha kerfuffle, brouhaha
not = covfefe
J.W. Brewer said,
October 30, 2021 @ 12:21 pm
To Scott Mauldin's point, the question then becomes how you assess or measure "average difficulty for a speaker of any world language." GIven an arbitrarily large amount of grant money, one could try to study the issue empirically by selecting a benchmark group of six or eight "starting" languages that differ significantly from each other but have a sufficient population of native speakers that (subject to the vagaries of the occasional coup or what have you) you could recruit demographically balanced samples of those populations as study participants and then have one sample group from each reference language try to learn the same L2 using some consistent metrics to assess how easy or hard it was. I don't have any strong intuitions about how tightly or loosely clustered the results would be for an L2 not strongly similar to any of the benchmark languages, but would love to have someone write up some grant proposals and get started.
My off-the-cuff suggestion for benchmark languages (I'm deliberately excluding English and certain other excessively-high-profile ones) is Amharic, Azerbaijani, Burmese, Korean, Malay, Portuguese, Telugu, and Yoruba. It would be easy to come up with other lists that don't include any of those but would approximate the same resulting range of starting points. (A range that by necessity only captures some subpart of the entire range of variation in all extant natural languages.) You might want to keep track of who in your test sample also had decent L2 grasp of English or some other major "world language," as a possible source of interference with clear results.
Not a naive speaker said,
October 30, 2021 @ 2:59 pm
The is a "maximum complexity" for a spoken language. If it were too complex it either would have died out, reduced in complexity or be the "Zeitvertreib" of a leisure class. The purpose of a spoken language is to communicate with other speakers.
Think about it: if MSM(spoken) were so terribly difficult, no speaker would use a telephone, just fax machines
Bob Michael said,
October 30, 2021 @ 6:56 pm
Certainly all languages are equally easy to learn from childhood as L1, but it appears that among the world’s dominant languages, none are polysynthetic. I wonder if a rapidly expanding language undergoes certain processes that make it easier to learn for new speakers? Some degree of simplification and grammar reduction occurs as new speakers adopt the language. So perhaps we can say that, beyond a certain point, languages with complex or unpredictable grammar are actually harder for most adults to learn?
Doug said,
October 31, 2021 @ 6:23 am
"Certainly all languages are equally easy to learn from childhood as L1…"
How can we be certain of that, give that many of the world's languages have not been thoroughly documented by linguists?
Stephen L said,
October 31, 2021 @ 4:24 pm
One thing I've wondered about (as a non-linguistic person) is a measure of what age people tend to reach proficiency in the 'adult'/'correct'/'standard' version of a language (while aware it's a sliding/fuzzy scale, and dialectical/sociological issues, generational differences in speech, and so on…) between language-speaking groups for different languages/countries.
Not sure where you'd find such data, though.
Bob Michael said,
October 31, 2021 @ 8:33 pm
Doug said: "Certainly all languages are equally easy to learn from childhood as L1…"
How can we be certain of that, give that many of the world's languages have not been thoroughly documented by linguists?
Fair point. But a people with slower language development would be at a severe disadvantage, and any such language may have changed or died a long time ago. I can’t imagine there are big differences in how long it takes for children to learn their own language, but I’d like to see some studies.
Andreas Johansson said,
November 1, 2021 @ 5:17 am
I've read (in popular sources) that Czech and Danish tend to be acquired by children more slowly than other European languages, supposedly due to complex phonology and sloppy articulation respectively. So perhaps studies on their acquistion would be the place to start looking?
~flow said,
November 1, 2021 @ 10:06 am
"Czech and Danish tend to be acquired by children more slowly than other European languages, supposedly due to complex phonology and sloppy articulation"
The classical take on the difficulty Danish presents even to native speakers is this skit: https://www.youtube.com/watch?v=s-mOy8VUEBk (from a Norwegian TV show called Uti Vår Hage)
Andrew said,
November 2, 2021 @ 1:37 pm
"One of the lines of evidence she used was indeed was psycholinguistic, that immigrants' kids' accents and language, among other things, resemble their peers not their parents."
In college I knew the oldest child in a family of four children which exhibited an interesting variety in accents in English.
The parents were both British, well-educated with (broadly speaking) RP accents who had moved to the US as adults, with children in tow, for career reasons.
The two older children had come to the US as young teenagers and by the time I met the oldest one 10+ years later at college in the US, he still had a very distinctively British accent. I'm sure there were some subtle American influences that would have been evident to a BrE speaker in Britain, and for example he almost always used US car-related terminology (trunk instead of boot, hood instead of bonnet, etc.), having learned to drive in the US, but his accent very clearly did not read as American to anyone in the US.
Conversely his very youngest sibling, who was about 10 years younger than him (and thus had come to the US at roughly ~3yo), spoke exactly the same as his peers. I met him at his oldest sibling’s wedding — had I not met him in that context, or only spoken to him over the phone, there’s no way I would have guessed he had British parents.
The only (noticeable) giveaway was his use of “Mum” instead of “Mom” to his mother, but I observed that even here he only said “Mum” when speaking directly to his mother and used “mom” when speaking (say) to me or other Americans. My guess would be that “Mum” had very much insisted on being called that, even though she had ended up raising her children in the US.
Bob Michael said,
November 3, 2021 @ 10:40 pm
A quick search turned up this article, which claims Danish children have a harder time learning their mother tongue.
https://theconversation.com/danish-children-struggle-to-learn-their-vowel-filled-language-and-this-changes-how-adult-danes-interact-161143
“There has been a longstanding debate within the language sciences about whether all languages are similarly complex and whether this might affect how people’s brains learn and process language. Our discovery about Danish challenges the idea that all native languages are equally easy to learn and use. Indeed, learning different languages from birth may lead to distinct and separate ways of processing those languages.”
I would never have predicted this, if true.
Robin said,
November 4, 2021 @ 1:55 am
"Namely, subconsciously or even consciously, L1 speakers often rebel against and essentially refuse to learn the language of their parents when the latter force / coerce them to do so."
This reminded me of some dear friends of mine. He's Brazilian, she's Argentinian, and both their children were born in an Anglophone country. Portuguese and Spanish are used equally in their home, but curiously their youngest child resisted any language other than English from shortly after birth, apparently screaming and crying if addressed in either of his parents' L1s. When I first met him, he was about 3 and he understood all 3 languages of course, but flatly refused to use anything other than English – even a simple 'hola' or 'bom dia' drew a filthy scowl and silence. His parents acquiesced, and did not force him to use either. In the last 2-3 years a noticeable shift has occurred – he's happy to reply in either Spanish or Portuguese and actively enjoys being my tutor when I ask him "how do you say X in Y?" or when we talk about the difference and similarities between Spanish, Portuguese and Italian. A win for the 'softly softly' approach, I feel.
Gokul Madhavan said,
November 4, 2021 @ 2:35 pm
This is a bit of a sidebar response to Daphne Preston-Kendal's opening comment (how I wish websites had margins for people to scribble on and exchange notes on tangential topics!):
naras in Sanskrit can be either singular or plural, depending on what nominal stem you're looking at. If you're using nara-, the more common form in Classical Sanskrit, then naras is the nominative singular. But if you're using nṛ-, a rarer but likely older form closer to its PIE ancestor *h₂nḗr, then naras is the nominative/accusative plural.
Scott Mauldin said,
November 5, 2021 @ 5:58 am
@J.W. Brewer – I wasn't suggesting that measuring "average difficulty" was a feasible goal nor that the result of such a study would have much practical utility. It was just an exercise in conceptualization. But if you were going to make such a measure, sure, your methodology seems perfectly valid.