I wrote a piece for work on quota systems and affirmative action in employment (“Fixing Our Model of Meritocracy”). It’s politics-related, but I did get to cite a really fun natural experiment and talk about quotas for the use of countering the availability heuristic.
This is a tangent, but since you mention the “good founders started [programming] at 13” meme, it’s a little bit relevant …
I find it deeply bizarre that there’s this idea today among some programmers that if you didn’t start programming in your early teens, you will never be good at programming. Why is this so bizarre? Because until very recently, there was no such thing as a programmer who started at a young age; and yet there were people who became good at programming.
Prior to the 1980s, most people who ended up as programmers didn’t have access to a computer until university, often not until graduate school. Even for university students, relatively unfettered access to a computer was an unusual exception, found only in extremely hacker-friendly cultures such as MIT.
Put another way: Donald Knuth probably didn’t use a computer until he was around 20. John McCarthy was born in 1927 and probably couldn’t have come near a computer until he was a professor, in his mid-20s. (And of course Alan Turing, Jack Good, or John von Neumann couldn’t have grown up with computers!)
(But all of them were mathematicians, and several of them physicists. Knuth, for one, was also a puzzle aficionado and a musician from his early years — two intellectual pursuits often believed to correlate with programming ability.)
In any event, it should be evident from the historical record that people who didn’t see a computer until adulthood could still become extremely proficient programmers and computer scientists.
I’ve heard some people defend the “you can’t be good unless you started early” meme by comparison with language acquisition. Humans generally can’t gain native-level fluency in a language unless they are exposed to it as young children. But language acquisition is a very specific developmental process that has evolved over thousands of generations, and occurs in a developmentally-critical period of very early childhood. Programming hasn’t been around that long, and there’s no reason to believe that a critical developmental period in early adolescence could have come into existence in the last few human generations.
So as far as I can tell, we should really treat the idea that you have to start early to become a good programmer as a defensive and prejudicial myth, a bit of tribal lore arising in a recent (and powerful) subculture — which has the effect of excluding and driving off people who would be perfectly capable of learning to code, but who are not members of that subculture.
Seems to me that using computers since your childhood is not necessary, but there is something which is necessary, and which is likely to be expressed in childhood as an interest in computer programming. And, as you mentioned, in the absence of computers, this something is likely to be expressed as an interest in mathematics or physics.
So the correct model is not “early programming causes great programmers”, but rather “X causes great programmers, and X causes early programming; therefore early programming correlates with great programmers”.
Starting early with programming is not strictly necessary… but these days when computers are almost everywhere and they are relatively cheap, not expressing any interest in programming during one’s childhood is an evidence this person is probably not meant to be a good programmer. (The only question is how strong this evidence is.)
Comparing with language acquisition is wrong… unless the comparison is true for mathematics. (Is there a research on this?) Again, the model “you need programming acquisition as a child” would be wrong, but the model “you need math acquisition as a child, and without this you later will not grok programming” might be correct.
the correct model is not “early programming causes great programmers”, but rather “X causes great programmers, and X causes early programming; therefore early programming correlates with great programmers”.
Yeah, I think this is explicitly the claim Paul Graham made, with X = “deep interest in technology”.
The problem with that is I think, at least with technology companies, the people who are really good technology founders have a genuine deep interest in technology. In fact, I’ve heard startups say that they did not like to hire people who had only started programming when they became CS majors in college. If someone was going to be really good at programming they would have found it on their own. Then if you go look at the bios of successful founders this is invariably the case, they were all hacking on computers at age 13.
This is a tangent, but since you mention the “good founders started [programming] at 13” meme, it’s a little bit relevant …
There is a rule of thumb that achieving exceptional mastery in any specific field requires 10,000 hours of practice. This seems to be true across fields, in classical musicians, chess players, sports players, scholars/academics etc… It’s a lot easier to meet that standard if you start from childhood. Note that people who make this claim in the computing field are talking about hackers, not professional programmers in a general sense. It’s very possible to become a productive programmer at any age.
Humans generally can’t gain native-level fluency in a language unless they are exposed to it as young children.
The only aspect of language with a critical period is accent. Adults commonly achieve fluency. In fact, adults learn a second language faster than children.
As far as I know, the degree to which second-language speakers can acquire native-like competence in domains other than phonetics is somewhat debated. Anecdotally, it’s a rare person who manages to never make a syntactic error that a native speaker wouldn’t make, and there are some aspects of language (I’m told that subjunctive in French and aspect in Slavic languages may be examples) that may be impossible to fully acquire for non-native speakers.
So I wouldn’t accept this theoretical assertion without further evidence; and for all practical purposes, the claim that you have to learn a language as a child in order to become perfect (in the sense of native-like) with it is true.
Not my downvotes, but you’re probably getting flak for just asserting stuff and then demanding evidence for the opposing side. A more mellow approach like “huh that’s funny I’ve always heard the opposite” would be better received.
Indeed, I probably expressed myself quite badly, because I don’t think what I meant to say is that outrageous: I heard the opposite, and anecdotally, it seems right—so I would have liked to see the (non-anecdotal) evidence against it. Perhaps I phrased it a bit harshly because what I was responding to was also just an unsubstantiated assertion (or, alternatively, a non-sequitur in that it dropped the “native-like” before fluency).
As far as I know, the degree to which second-language speakers can acquire native-like competence in domains other than phonetics is somewhat debated.
Links? As far as I know it’s not debated.
there are some aspects of language (I’m told that subjunctive in French and aspect in Slavic languages may be examples) that may be impossible to fully acquire for non-native speakers.
That’s, ahem, bullshit. Why in the world would some features of syntax be “impossible to fully acquire”?
for all practical purposes, the claim that you have to learn a language as a child in order to become perfect (in the sense of native-like) with it is true.
You may easily know more about this issue than me, because I haven’t actually researched this.
That said, let’s be more precise. If we’re talking about mere fluency, there is, of course, no question.
But if we’re talking about actually native-equivalent competence and performance, I have severe doubts that this is even regularly achieved. How many L2 speakers of English do you know who never, ever pick an unnatural choice from among the myriad of different ways in which the future can be expressed in English? This is something that is completely effortless for native speakers, but very hard for L2 speakers.
The people I know who are candidates for that level of proficiency in an L2 are at the upper end of the intelligence spectrum, and I also know a non-dumb person who has lived in a German-speaking country for decades and still uses wrong plural formations. Hell, there’s people who are employed and teach at MIT and so are presumably non-dumb who say things like “how it sounds like”.
The two things I mentioned are semantic/pragmatic, not syntactic. I know there is a study that shows L2 learners don’t have much of a problem with the morphosyntax of Russian aspect, and that doesn’t surprise me very much. I don’t know and didn’t find any work that tried to test native-like performance on the semantic and pragmatic level.
I’m not sure how to answer the “why” question. Why should there be a critical period for anything? … Intuitively, I find that semantics/pragmatics, having to do with categorisation, is a better candidate for something critical-period-like than pure (morpho)syntax. I’m not even sure you need critical periods for everything, anyway. If A learns to play the piano starting at age 5 and B starts at age 35, I wouldn’t be surprised if A is not only on average, but almost always, better at age 25 than B is at 55. Unfortunately, that’s basically impossible to study while controlling for all confounders like general intelligence, quality of instruction, and number of hours spent on practice. (The piano example would be analogous more to the performance than the competence aspect of language, I suppose.)
There is a study about Russian dative subjects that suggests even highly advanced L2 speakers with lots of exposure don’t get things quite right. Admittedly, you can still complain that they don’t separate the people who have lived in a Russian-speaking country for only a couple of months from those who have lived there for a decade.
The thing about the subjunctive is, at best, wrong, but certainly not bullshit. The fact that it was told to me by a very intelligent French linguist about a friend of his whose L2-French is flawless except for occasional errors in that domain is better evidence for that being a very hard thing to acquire than your “bullshit” is against that.
How many L2 speakers of English do you know who never, ever pick an unnatural choice from among the myriad of different ways in which the future can be expressed in English?
You are committing the nirvana fallacy. How many native speakers of English never make mistakes or never “pick an unnatural choice”?
For example, I know a woman who immigrated to the US as an adult and is fully bilingual. As an objective measure, I think she had the perfect score on the verbal section of the LSAT. She speaks better English than most “natives”. She is not unusual.
The fact that it was told to me by a very intelligent French linguist about a friend of his whose L2-French is flawless except for occasional errors in that domain
Tell your French linguist to go into countryside and listen to the French of the uneducated native speakers. Do they make mistakes?
How many native speakers of English never make mistakes or never “pick an unnatural choice”?
I’m not talking about performance errors in general. I’m talking about the fact that it is extremely hard to acquire native-like competence wrt the semantics and pragmatics of the ways in which English allows one to express something about the future.
She speaks better English than most “natives”.
Your utterance of this sentence severely damages your credibility with respect to any linguistic issue. The proper way to say this is: she speaks higher-status English than most native speakers. Besides, the fact that she gets perfect scores on some test (whose content and format is unknown to me), which presumably native speakers don’t, suggests that she is far from an average individual anyway.
Also, that you’re not bringing up a single relevant study that compares long-time L2 speakers with native speakers on some interesting, intricate and subtle issue where a competence difference might be suspected leaves me with a very low expectation of the fruitfulness of this discussion, so maybe we should just leave it at that. I’m not even sure to what extent we aren’t simply talking past each other because we have different ideas about what native-like performance means.
Tell your French linguist to go into countryside and listen to the French of the uneducated native speakers. Do they make syntax errors?
They don’t, by definition; not the way you probably mean it. I wouldn’t know why the rate of performance errors should correlate in any way with education (controlling for intelligence). I also trust the man’s judgment enough to assume that he was talking about a sort of error that stuck out because a native speaker wouldn’t make it.
I’m talking about the fact that it is extremely hard to acquire native-like competence wrt the semantics and pragmatics of the ways in which English allows one to express something about the future.
I don’t think so. This looks like an empirical question—what do you mean by “extremely hard”? Any evidence?
Your utterance of this sentence severely damages your credibility with respect to any linguistic issue. The proper way to say this is: she speaks higher-status English than most native speakers.
No, I still don’t think so—for either of your claims. Leaving aside my credibility, non-black English in the United States (as opposed to the UK) has few ways to show status and they tend to be regional, anyway. She speaks better English (with some accent, to be sure) in the usual sense—she has a rich vocabulary and doesn’t make many mistakes.
she is far from an average individual anyway.
While that is true, your claims weren’t about averages. Your claims were about impossibility—for anyone. An average person isn’t successful at anything, including second languages.
I don’t think so. This looks like an empirical question—what do you mean by “extremely hard”? Any evidence?
I don’t know if anybody has ever studied this—I would be surprised if they had -, so I have only anecdotal evidence from the uncertainty I myself experience sometimes when choosing between “will”, “going to”, plain present, “will + progressive”, and present progressive, and from the testimony of other highly advanced L2 speakers I’ve talked to who feel the same way—while native speakers are usually not even aware that there is an issue here.
She speaks better English (with some accent, to be sure) in the usual sense—she has a rich vocabulary and doesn’t make many mistakes.
How exactly is “rich vocabulary” not high-status? (Also, are you sure it actually contains more non-technical lexemes and not just higher-status lexemes?) I’m not exactly sure what you mean by “mistakes”. Things that are ungrammatical in your idiolect of English?
While that is true, your claims weren’t about averages. Your claims were about impossibility—for anyone. An average person isn’t successful at anything, including second languages.
I actually made two claims. The one was that it’s not entirely clear that there aren’t any such in-principle impossibilities, though I admit that the case for them isn’t very strong. I will be very happy if you give me a reference surveying some research on this and saying that the empirical side is really settled and the linguists who still go on telling their students that it isn’t are just not up-to-date.
The second is that in any case, only the most exceptional L2 learners can in practice expect to ever achieve native-like fluency.
It seems you are talking about being self-conscious, not about language fluency.
I didn’t say it was about fluency. But I don’t think it’s about self-consciousness, either. Native speakers of a language pick the appropriate tense and aspect forms of verbs perfectly effortlessly—or how often do you hear a native speaker of English use a progressive in a case where it strikes you as inappropriate and you would say that they should really have used a plain tense here, for example?* - while for L2 speakers, it is generally pretty hard to grasp all the details of a language’s tense/aspect system.
*I’m choosing the progressive as an example because it’s easiest to describe, not because I think it’s a candidate for serious unacquirability. It’s known to be quite hard for native speakers of a language that has no aspect, but it’s certainly possible to get to a point where you don’t use the progressive wrongly essentially ever.
What possible mechanism do you have in mind?
For syntax, you would really need to be a strong Chomskian to expect any such things. For semantics, it seems to be a bit more plausible a priori: maybe as an adult, you have a hard time learning new ways of carving up the world?
Well, let’s get specific. Which test do you assert native speakers will pass and ESL people will not (except for the “most exceptional”)?
I don’t know of a pass/fail format test, but I expect reading speed and the speed of their speech to be lower in L2 speakers than in L1 speakers of comparable intelligence. I would also expect that if you measure cognitive load somehow, language processing in an L2 requires more of your capacity than processing your L1. I would also expect that the active vocabulary of L1 speakers is generally larger than that of an L2 speaker even if all the words in the L1 speaker’s active lexicon are in the L2 speaker’s passive vocabulary.
The things being measured are different. To a first
approximation, all native speakers do maximally well at
sounding like a native speaker.
Lumifer’s
friend
may indeed speak like a native speaker (though it’s rare for
people who learned as adults to do so), but she cannot be
better at it than “most ‘natives’”.
Or maybe it means that high status and low status English have different difficulties, and native speakers tend to learn the one that their parents use (finding others harder) while L2 speakers learn to speak from a description of English which is actually a description of a particular high status accent (usually either Oxford or New England I think)
The “Standard American Accent” spoken in the media and generally taught to foriegners is the confusingly named “Midwestern” Accent, which due to internal migration and a subsequent vowel shift, is now mostly spoken in California and the Pacific Northwest.
Interestingly enough, my old Japanese instructor was a native Osakan, who’s natural dialect was Kansai-ben; despite this, she conducted the class using the standard, Tokyo Dialect.
If A learns to play the piano starting at age 5 and B starts at age 35, I wouldn’t be surprised if A is not only on average, but almost always, better at age 25 than B is at 55. Unfortunately, that’s basically impossible to study while controlling for all confounders like general intelligence, quality of instruction, and number of hours spent on practice.
If all you are saying is that people who start learning a language at age 2 are almost always better at it than people who start learning the same language at age 20, I don’t think anyone would disagree. The whole discussion is about controlling for confounders...
Yes and no—the whole discussion is actually two discussions, I think.
One is about in-principle possibility, the presence of something like a critical period, etc. There it is crucial for confounders.
The second discussion is about in-practice possibility, whether people starting later can reasonably expect to get to the same level of proficiency. Here the “confounders” are actually part of what this is about.
There is a study about Russian dative subjects that suggests even highly advanced L2 speakers with lots of exposure don’t get things quite right.
Bonus points for giving a specific example, which helped me to understand your point, and at this moment I fully agree with you. Because I understand the example; my own language has something similar, and wouldn’t expect a stranger to use this correctly. The reason is that it would be too much work to learn properly, for too little benefit. It’s a different way to say things, and you only achieve a small difference in meaning. And even if you asked a non-linguist native, they would probably find it difficult to explain the difference properly. So you have little chance to learn it right, and also little motivation to do.
Here is my attempt to explain the examples from the link, pages 3 and 4. (I am not a Russian language speaker, but my native language is also Slavic, and I learned Russian. If I got something wrong, please correct me.)
That’s pretty much the same meaning, it’s just that the first variant is “more agenty”, and the second variant is “less agenty”, to use the LW lingo. But that’s kinda difficult to explain explicitly, becase… you know, how exactly can “hearing” (not active listening, just hearing) be “agenty”; and how exactly can “wanting” be “non-agenty”? It doesn’t seem to make much sense, until you think about it, right? (The “non-agenty wanting” is something like: my emotions made me to want. So I admit that I wanted, but at the same time I deny full responsibility for my wanting.)
As a stranger, what is the chance that (1) you will hear it explained in a way that will make sense to you, (2) you will remember it correctly, and (3) when the opportunity comes, you will remember to use it. Pretty much zero, I guess. Unless you decide to put an extra effort into this aspect of the langauge specifically. But considering the costs and benefits, you are extremely unlikely to do it, unless being a professional translator to Russian is extremely important for you. (Or unless you speak a Slavic language that has a similar concept, so the costs are lower for you, but even then you need a motivation to be very good at Russian.)
Now when you think about contexts, these kinds of words are likely to be used in stories, but don’t appear in technical literature or official documents, etc. So if you are a Russian child, you heard them a lot. If you are a Russian-speaking foreigner working in Russia, there is a chance you will literally never hear it at the workplace.
The paper doesn’t even find a statistically significant difference. The point estimate is that advanced L2 do worse than natives, but natives make almost as many mistakes.
They did found differences with the advances L2 speakers, but I guess we care about the highly advanced ones. They point out a difference at the bottom of page 18, though admittedly, it doesn’t seem to be that much of a big deal and I don’t know enough about statistics to tell whether it’s very meaningful.
Ah I see, yes you are right. That is the correct plural in this case. Sorry about that! ‘Mne poslyshalos chtoto’ (“something made itself heard by me”) would be the singular, vs the plural above (“the steps on the roof made themselves heard by me.”). Or at least I think it would be—I might be losing my ear for Russian.
Huh? What a curious misunderstanding! The theoretical referred just the—theoretical! - question of whether it’s in principle possible to acquire native-like proficiency, which was contrasted with my claim that even if it is, most people cannot expect to reach that state in practice.
You really think it’s common for L2 speakers to achieve native-like levels of proficiency? Where do you live and who are these geniuses? I’m serious. For example, I see people speaking at conferences who have lived in the US for years, but aren’t native speakers, and they are still not doing so with native-like fluency and eloquence. And presumably you have to be more than averagely intelligent to give a talk at a scientific conference...
I’m not talking about just any kind of fluency here, and neither was fubarobfusco, I assume. I suspect I was trying to interpret your utterance in a way that I didn’t assign very low probability to (i.e. not as claiming that it’s common for people to become native-like) and that also wasn’t a non-sequitur wrt the claim you were referring to (by reducing native-like fluency to some weaker notion) and kind of failed.
Maybe I should have said “routinely” rather than “commonly.” But the key differentiator is effort.
I don’t care about your theoretical question of whether you can come up with a test that L2 speakers fail. I assume that fubarobfusco meant the same thing I meant. I’m done.
I find it deeply bizarre that there’s this idea today among some programmers that if you didn’t start programming in your early teens, you will never be good at programming.
Suppose you replaced it with the idea that people who started programming when they were 13 have a much easier time becoming good programmers as adults, and so are overrepresented among programmers at every level. Does that still sound bizarre?
The same tortured analysis plays out in the business world, where Paul Graham, the head of YCombinator, a startup incubator, explained that one reason his company funds fewer women-led companies is because fewer of them fit this profile of a successful founder:
If someone was going to be really good at programming they would have found it on their own. Then if you go look at the bios of successful founders this is invariably the case, they were all hacking on computers at age 13.
The trouble is, successful founders don’t run through a pure meritocracy, either. They’re supported, mentored, and funded when they’re chosen by venture capitalists like Graham. And, if everyone is working on the same model of “good founders started at 13″ then a lot of clever ideas, created by people of either gender, might get left on the table.
But even if the government were keeping better tabs on affirmative action, the bigger problem is that its jurisdiction doesn’t reach the parts of the economy where affirmative action is most desperately needed: the places where real money is made and real power is allocated. The best example of this is the industry that dominates so much of our economy today: the technology sector. Silicon Valley’s racial diversity is pretty terrible, the kind of gross imbalance that inspires special reports on CNN.
It’s a dismal state of affairs, but how could it really be otherwise? Silicon Valley isn’t just an industry; it’s a social and cultural ecosystem that grew out of a very specific social and cultural setting: mostly West Coast, upper-middle-class white guys who liked to tinker with motherboards and microchips. If you were around that culture, you became a part of it. If you weren’t, you didn’t. And because of the social segregation that pervades our society, very few black people were around to be a part of it.
Some would purport to remedy this by fixing the tech industry job pipeline: more STEM graduates, more minority internships and boot camps, etc. And that will get you changes here and there, at the margins, but it doesn’t get at the real problem. The big success stories of the Internet age—Instagram, YouTube, Twitter—all came about in similar ways: A couple of people had an idea, they got together with some of their friends, built something, called some other friends who knew some other friends who had access to friends with serious money, and then the thing took off and now we’re all using it and they’re all millionaires. The process is organic, somewhat accidental, and it moves really, really fast. And by the time those companies are big enough to worry about their “diversity,” the ground-floor opportunities have already been spoken for
I almost upvoted your post on realizing you are a woman and thinking I’d like more women on LW. Then I realized how ironic that was. Then I did it anyway, likely influenced by the pretty photo on your article (not caring whether it was stock or you).
Fixing our meritocracy presumes we have a meritocracy to fix. Certainly a democracy is not a meritocracy, unless your definition of merit is EXTREMELY flexible to the point of defining merit as “getting elected.” Certainly the Athenian and 19th century American democracies which supported human slavery were not meritocracies, unless again your definition of merit is flexible enough to include being white and/or patrician.
If there is a lesson from the study you cite, it would seem to be that one should push for quotas at the governmental level. It is said by many, but I don’t know the evidence, that an advantage Europe and US have over Islamic societies is that we are much better about monetizing the talents of women, and so we have up to 200% more per capita effective productivity available. Your Indian lesson shows an example where rather extreme and anti-democratic quotas appeared to shift the preferences of the broad population to include more humans more broadly in what they see as the talent pool.
Is it likely that quotas in the US have worked negatively rather than positively? Looked at myopically one might make the case. But pre-quota US was a MUCH LESS integrated society. I grew up in a middle class suburb in Long Island (Farmingdale) hardly a bastion of white privelege. In the 1970s, a black family bought a house and had stuff thrown through their windows and a range of other harassments perpetrated upon them by anonymous but I’m willing to bet white perpetrators. Now we have interracial couples all over the southern US, and tremendous reduction in racist feeling in people younger than myself. Correlation is not causation, but it ain’t exactly an argument against causation either.
I wrote a piece for work on quota systems and affirmative action in employment (“Fixing Our Model of Meritocracy”). It’s politics-related, but I did get to cite a really fun natural experiment and talk about quotas for the use of countering the availability heuristic.
This is a tangent, but since you mention the “good founders started [programming] at 13” meme, it’s a little bit relevant …
I find it deeply bizarre that there’s this idea today among some programmers that if you didn’t start programming in your early teens, you will never be good at programming. Why is this so bizarre? Because until very recently, there was no such thing as a programmer who started at a young age; and yet there were people who became good at programming.
Prior to the 1980s, most people who ended up as programmers didn’t have access to a computer until university, often not until graduate school. Even for university students, relatively unfettered access to a computer was an unusual exception, found only in extremely hacker-friendly cultures such as MIT.
Put another way: Donald Knuth probably didn’t use a computer until he was around 20. John McCarthy was born in 1927 and probably couldn’t have come near a computer until he was a professor, in his mid-20s. (And of course Alan Turing, Jack Good, or John von Neumann couldn’t have grown up with computers!)
(But all of them were mathematicians, and several of them physicists. Knuth, for one, was also a puzzle aficionado and a musician from his early years — two intellectual pursuits often believed to correlate with programming ability.)
In any event, it should be evident from the historical record that people who didn’t see a computer until adulthood could still become extremely proficient programmers and computer scientists.
I’ve heard some people defend the “you can’t be good unless you started early” meme by comparison with language acquisition. Humans generally can’t gain native-level fluency in a language unless they are exposed to it as young children. But language acquisition is a very specific developmental process that has evolved over thousands of generations, and occurs in a developmentally-critical period of very early childhood. Programming hasn’t been around that long, and there’s no reason to believe that a critical developmental period in early adolescence could have come into existence in the last few human generations.
So as far as I can tell, we should really treat the idea that you have to start early to become a good programmer as a defensive and prejudicial myth, a bit of tribal lore arising in a recent (and powerful) subculture — which has the effect of excluding and driving off people who would be perfectly capable of learning to code, but who are not members of that subculture.
Seems to me that using computers since your childhood is not necessary, but there is something which is necessary, and which is likely to be expressed in childhood as an interest in computer programming. And, as you mentioned, in the absence of computers, this something is likely to be expressed as an interest in mathematics or physics.
So the correct model is not “early programming causes great programmers”, but rather “X causes great programmers, and X causes early programming; therefore early programming correlates with great programmers”.
Starting early with programming is not strictly necessary… but these days when computers are almost everywhere and they are relatively cheap, not expressing any interest in programming during one’s childhood is an evidence this person is probably not meant to be a good programmer. (The only question is how strong this evidence is.)
Comparing with language acquisition is wrong… unless the comparison is true for mathematics. (Is there a research on this?) Again, the model “you need programming acquisition as a child” would be wrong, but the model “you need math acquisition as a child, and without this you later will not grok programming” might be correct.
Yeah, I think this is explicitly the claim Paul Graham made, with X = “deep interest in technology”.
There is a rule of thumb that achieving exceptional mastery in any specific field requires 10,000 hours of practice. This seems to be true across fields, in classical musicians, chess players, sports players, scholars/academics etc… It’s a lot easier to meet that standard if you start from childhood. Note that people who make this claim in the computing field are talking about hackers, not professional programmers in a general sense. It’s very possible to become a productive programmer at any age.
The only aspect of language with a critical period is accent. Adults commonly achieve fluency. In fact, adults learn a second language faster than children.
As far as I know, the degree to which second-language speakers can acquire native-like competence in domains other than phonetics is somewhat debated. Anecdotally, it’s a rare person who manages to never make a syntactic error that a native speaker wouldn’t make, and there are some aspects of language (I’m told that subjunctive in French and aspect in Slavic languages may be examples) that may be impossible to fully acquire for non-native speakers.
So I wouldn’t accept this theoretical assertion without further evidence; and for all practical purposes, the claim that you have to learn a language as a child in order to become perfect (in the sense of native-like) with it is true.
Not my downvotes, but you’re probably getting flak for just asserting stuff and then demanding evidence for the opposing side. A more mellow approach like “huh that’s funny I’ve always heard the opposite” would be better received.
Indeed, I probably expressed myself quite badly, because I don’t think what I meant to say is that outrageous: I heard the opposite, and anecdotally, it seems right—so I would have liked to see the (non-anecdotal) evidence against it. Perhaps I phrased it a bit harshly because what I was responding to was also just an unsubstantiated assertion (or, alternatively, a non-sequitur in that it dropped the “native-like” before fluency).
[error]
Links? As far as I know it’s not debated.
That’s, ahem, bullshit. Why in the world would some features of syntax be “impossible to fully acquire”?
For all practical purposes it is NOT true.
You may easily know more about this issue than me, because I haven’t actually researched this.
That said, let’s be more precise. If we’re talking about mere fluency, there is, of course, no question.
But if we’re talking about actually native-equivalent competence and performance, I have severe doubts that this is even regularly achieved. How many L2 speakers of English do you know who never, ever pick an unnatural choice from among the myriad of different ways in which the future can be expressed in English? This is something that is completely effortless for native speakers, but very hard for L2 speakers.
The people I know who are candidates for that level of proficiency in an L2 are at the upper end of the intelligence spectrum, and I also know a non-dumb person who has lived in a German-speaking country for decades and still uses wrong plural formations. Hell, there’s people who are employed and teach at MIT and so are presumably non-dumb who say things like “how it sounds like”.
The two things I mentioned are semantic/pragmatic, not syntactic. I know there is a study that shows L2 learners don’t have much of a problem with the morphosyntax of Russian aspect, and that doesn’t surprise me very much. I don’t know and didn’t find any work that tried to test native-like performance on the semantic and pragmatic level.
I’m not sure how to answer the “why” question. Why should there be a critical period for anything? … Intuitively, I find that semantics/pragmatics, having to do with categorisation, is a better candidate for something critical-period-like than pure (morpho)syntax. I’m not even sure you need critical periods for everything, anyway. If A learns to play the piano starting at age 5 and B starts at age 35, I wouldn’t be surprised if A is not only on average, but almost always, better at age 25 than B is at 55. Unfortunately, that’s basically impossible to study while controlling for all confounders like general intelligence, quality of instruction, and number of hours spent on practice. (The piano example would be analogous more to the performance than the competence aspect of language, I suppose.)
There is a study about Russian dative subjects that suggests even highly advanced L2 speakers with lots of exposure don’t get things quite right. Admittedly, you can still complain that they don’t separate the people who have lived in a Russian-speaking country for only a couple of months from those who have lived there for a decade.
The thing about the subjunctive is, at best, wrong, but certainly not bullshit. The fact that it was told to me by a very intelligent French linguist about a friend of his whose L2-French is flawless except for occasional errors in that domain is better evidence for that being a very hard thing to acquire than your “bullshit” is against that.
You are committing the nirvana fallacy. How many native speakers of English never make mistakes or never “pick an unnatural choice”?
For example, I know a woman who immigrated to the US as an adult and is fully bilingual. As an objective measure, I think she had the perfect score on the verbal section of the LSAT. She speaks better English than most “natives”. She is not unusual.
Tell your French linguist to go into countryside and listen to the French of the uneducated native speakers. Do they make mistakes?
I’m not talking about performance errors in general. I’m talking about the fact that it is extremely hard to acquire native-like competence wrt the semantics and pragmatics of the ways in which English allows one to express something about the future.
Your utterance of this sentence severely damages your credibility with respect to any linguistic issue. The proper way to say this is: she speaks higher-status English than most native speakers. Besides, the fact that she gets perfect scores on some test (whose content and format is unknown to me), which presumably native speakers don’t, suggests that she is far from an average individual anyway.
Also, that you’re not bringing up a single relevant study that compares long-time L2 speakers with native speakers on some interesting, intricate and subtle issue where a competence difference might be suspected leaves me with a very low expectation of the fruitfulness of this discussion, so maybe we should just leave it at that. I’m not even sure to what extent we aren’t simply talking past each other because we have different ideas about what native-like performance means.
They don’t, by definition; not the way you probably mean it. I wouldn’t know why the rate of performance errors should correlate in any way with education (controlling for intelligence). I also trust the man’s judgment enough to assume that he was talking about a sort of error that stuck out because a native speaker wouldn’t make it.
I don’t think so. This looks like an empirical question—what do you mean by “extremely hard”? Any evidence?
No, I still don’t think so—for either of your claims. Leaving aside my credibility, non-black English in the United States (as opposed to the UK) has few ways to show status and they tend to be regional, anyway. She speaks better English (with some accent, to be sure) in the usual sense—she has a rich vocabulary and doesn’t make many mistakes.
While that is true, your claims weren’t about averages. Your claims were about impossibility—for anyone. An average person isn’t successful at anything, including second languages.
I don’t know if anybody has ever studied this—I would be surprised if they had -, so I have only anecdotal evidence from the uncertainty I myself experience sometimes when choosing between “will”, “going to”, plain present, “will + progressive”, and present progressive, and from the testimony of other highly advanced L2 speakers I’ve talked to who feel the same way—while native speakers are usually not even aware that there is an issue here.
How exactly is “rich vocabulary” not high-status? (Also, are you sure it actually contains more non-technical lexemes and not just higher-status lexemes?) I’m not exactly sure what you mean by “mistakes”. Things that are ungrammatical in your idiolect of English?
I actually made two claims. The one was that it’s not entirely clear that there aren’t any such in-principle impossibilities, though I admit that the case for them isn’t very strong. I will be very happy if you give me a reference surveying some research on this and saying that the empirical side is really settled and the linguists who still go on telling their students that it isn’t are just not up-to-date.
The second is that in any case, only the most exceptional L2 learners can in practice expect to ever achieve native-like fluency.
It seems you are talking about being self-conscious, not about language fluency.
Why in the world would there be “in-principle impossibilities”—where does this idea even come from? What possible mechanism do you have in mind?
Well, let’s get specific. Which test do you assert native speakers will pass and ESL people will not (except for the “most exceptional”)?
I didn’t say it was about fluency. But I don’t think it’s about self-consciousness, either. Native speakers of a language pick the appropriate tense and aspect forms of verbs perfectly effortlessly—or how often do you hear a native speaker of English use a progressive in a case where it strikes you as inappropriate and you would say that they should really have used a plain tense here, for example?* - while for L2 speakers, it is generally pretty hard to grasp all the details of a language’s tense/aspect system.
*I’m choosing the progressive as an example because it’s easiest to describe, not because I think it’s a candidate for serious unacquirability. It’s known to be quite hard for native speakers of a language that has no aspect, but it’s certainly possible to get to a point where you don’t use the progressive wrongly essentially ever.
For syntax, you would really need to be a strong Chomskian to expect any such things. For semantics, it seems to be a bit more plausible a priori: maybe as an adult, you have a hard time learning new ways of carving up the world?
I don’t know of a pass/fail format test, but I expect reading speed and the speed of their speech to be lower in L2 speakers than in L1 speakers of comparable intelligence. I would also expect that if you measure cognitive load somehow, language processing in an L2 requires more of your capacity than processing your L1. I would also expect that the active vocabulary of L1 speakers is generally larger than that of an L2 speaker even if all the words in the L1 speaker’s active lexicon are in the L2 speaker’s passive vocabulary.
I wonder if there’s an implication that colloquial language is more complex than high status language.
The things being measured are different. To a first approximation, all native speakers do maximally well at sounding like a native speaker.
Lumifer’s friend may indeed speak like a native speaker (though it’s rare for people who learned as adults to do so), but she cannot be better at it than “most ‘natives’”.
What she can be better at than most natives is:
Vocabulary.
Speaking a high-status dialect (e.g., avoiding third person singular “don’t”, double negatives, and “there’s” + plural).
Using complex sentence structures.
Avoiding disfluencies.
It is possible, though, for a lower-status dialect to be more complex than a higher-status one. Example: the Black English verb system.
Or maybe it means that high status and low status English have different difficulties, and native speakers tend to learn the one that their parents use (finding others harder) while L2 speakers learn to speak from a description of English which is actually a description of a particular high status accent (usually either Oxford or New England I think)
The “Standard American Accent” spoken in the media and generally taught to foriegners is the confusingly named “Midwestern” Accent, which due to internal migration and a subsequent vowel shift, is now mostly spoken in California and the Pacific Northwest.
Interestingly enough, my old Japanese instructor was a native Osakan, who’s natural dialect was Kansai-ben; despite this, she conducted the class using the standard, Tokyo Dialect.
If all you are saying is that people who start learning a language at age 2 are almost always better at it than people who start learning the same language at age 20, I don’t think anyone would disagree. The whole discussion is about controlling for confounders...
Yes and no—the whole discussion is actually two discussions, I think.
One is about in-principle possibility, the presence of something like a critical period, etc. There it is crucial for confounders.
The second discussion is about in-practice possibility, whether people starting later can reasonably expect to get to the same level of proficiency. Here the “confounders” are actually part of what this is about.
Bonus points for giving a specific example, which helped me to understand your point, and at this moment I fully agree with you. Because I understand the example; my own language has something similar, and wouldn’t expect a stranger to use this correctly. The reason is that it would be too much work to learn properly, for too little benefit. It’s a different way to say things, and you only achieve a small difference in meaning. And even if you asked a non-linguist native, they would probably find it difficult to explain the difference properly. So you have little chance to learn it right, and also little motivation to do.
Here is my attempt to explain the examples from the link, pages 3 and 4. (I am not a Russian language speaker, but my native language is also Slavic, and I learned Russian. If I got something wrong, please correct me.)
“ya uslyshala …” = “I heard …”
″mne poslyshalis …” = “to-me happened-to-be-heard …”
“ya xotel …” = “I wanted …”
″mne xotelos …” = “to-me happened-to-want …”
That’s pretty much the same meaning, it’s just that the first variant is “more agenty”, and the second variant is “less agenty”, to use the LW lingo. But that’s kinda difficult to explain explicitly, becase… you know, how exactly can “hearing” (not active listening, just hearing) be “agenty”; and how exactly can “wanting” be “non-agenty”? It doesn’t seem to make much sense, until you think about it, right? (The “non-agenty wanting” is something like: my emotions made me to want. So I admit that I wanted, but at the same time I deny full responsibility for my wanting.)
As a stranger, what is the chance that (1) you will hear it explained in a way that will make sense to you, (2) you will remember it correctly, and (3) when the opportunity comes, you will remember to use it. Pretty much zero, I guess. Unless you decide to put an extra effort into this aspect of the langauge specifically. But considering the costs and benefits, you are extremely unlikely to do it, unless being a professional translator to Russian is extremely important for you. (Or unless you speak a Slavic language that has a similar concept, so the costs are lower for you, but even then you need a motivation to be very good at Russian.)
Now when you think about contexts, these kinds of words are likely to be used in stories, but don’t appear in technical literature or official documents, etc. So if you are a Russian child, you heard them a lot. If you are a Russian-speaking foreigner working in Russia, there is a chance you will literally never hear it at the workplace.
The paper doesn’t even find a statistically significant difference. The point estimate is that advanced L2 do worse than natives, but natives make almost as many mistakes.
They did found differences with the advances L2 speakers, but I guess we care about the highly advanced ones. They point out a difference at the bottom of page 18, though admittedly, it doesn’t seem to be that much of a big deal and I don’t know enough about statistics to tell whether it’s very meaningful.
‘mne poslyshalos’ I think. This one has connotations of ‘hearing things,’ though.
Note: “Mne poslyshalis’ shagi na krishe.” was the original example; I just removed the unchanging parts of the sentences.
Ah I see, yes you are right. That is the correct plural in this case. Sorry about that! ‘Mne poslyshalos chtoto’ (“something made itself heard by me”) would be the singular, vs the plural above (“the steps on the roof made themselves heard by me.”). Or at least I think it would be—I might be losing my ear for Russian.
What do you mean by “theoretical”? Is this just an insult you fling at people you disagree with?
Huh? What a curious misunderstanding! The theoretical referred just the—theoretical! - question of whether it’s in principle possible to acquire native-like proficiency, which was contrasted with my claim that even if it is, most people cannot expect to reach that state in practice.
I thought that my choice of the word “commonly” indicated that I was not talking about the limits of the possible.
You really think it’s common for L2 speakers to achieve native-like levels of proficiency? Where do you live and who are these geniuses? I’m serious. For example, I see people speaking at conferences who have lived in the US for years, but aren’t native speakers, and they are still not doing so with native-like fluency and eloquence. And presumably you have to be more than averagely intelligent to give a talk at a scientific conference...
I’m not talking about just any kind of fluency here, and neither was fubarobfusco, I assume. I suspect I was trying to interpret your utterance in a way that I didn’t assign very low probability to (i.e. not as claiming that it’s common for people to become native-like) and that also wasn’t a non-sequitur wrt the claim you were referring to (by reducing native-like fluency to some weaker notion) and kind of failed.
Maybe I should have said “routinely” rather than “commonly.” But the key differentiator is effort.
I don’t care about your theoretical question of whether you can come up with a test that L2 speakers fail. I assume that fubarobfusco meant the same thing I meant. I’m done.
Suppose you replaced it with the idea that people who started programming when they were 13 have a much easier time becoming good programmers as adults, and so are overrepresented among programmers at every level. Does that still sound bizarre?
Donald Knuth was probably doing real math in his early teens. Maybe this counts.
A similar argument was presented in an article at Slate: Affirmative action doesn’t work. It never did. It’s time for a new solution.:
I almost upvoted your post on realizing you are a woman and thinking I’d like more women on LW. Then I realized how ironic that was. Then I did it anyway, likely influenced by the pretty photo on your article (not caring whether it was stock or you).
Fixing our meritocracy presumes we have a meritocracy to fix. Certainly a democracy is not a meritocracy, unless your definition of merit is EXTREMELY flexible to the point of defining merit as “getting elected.” Certainly the Athenian and 19th century American democracies which supported human slavery were not meritocracies, unless again your definition of merit is flexible enough to include being white and/or patrician.
If there is a lesson from the study you cite, it would seem to be that one should push for quotas at the governmental level. It is said by many, but I don’t know the evidence, that an advantage Europe and US have over Islamic societies is that we are much better about monetizing the talents of women, and so we have up to 200% more per capita effective productivity available. Your Indian lesson shows an example where rather extreme and anti-democratic quotas appeared to shift the preferences of the broad population to include more humans more broadly in what they see as the talent pool.
Is it likely that quotas in the US have worked negatively rather than positively? Looked at myopically one might make the case. But pre-quota US was a MUCH LESS integrated society. I grew up in a middle class suburb in Long Island (Farmingdale) hardly a bastion of white privelege. In the 1970s, a black family bought a house and had stuff thrown through their windows and a range of other harassments perpetrated upon them by anonymous but I’m willing to bet white perpetrators. Now we have interracial couples all over the southern US, and tremendous reduction in racist feeling in people younger than myself. Correlation is not causation, but it ain’t exactly an argument against causation either.