by a Nobel laureate in physics who is a native speaker of Dutch, makes clear what the key learning task is to be a good physicist: "English is a prerequisite. If you haven't mastered it yet, learn it. You must be able to read, write, speak and understand English." On his list of things to learn for physics, that even comes before mathematics.
I like to share advice on language learning, because this topic comes up on Hacker News frequently. I hope the FAQ information below helps hackers achieve their dreams. As I learned Mandarin Chinese up to the level that I was able to support my family for several years as a Chinese-English translator and interpreter, I had to tackle several problems for which there is not yet a one-stop-shopping software solution. For ANY pair of languages, even closely cognate pairs of West Germanic languages like English and Dutch, or Wu Chinese dialects like those of Shanghai and Suzhou, the two languages differ in sound system, so that what is a phoneme in one language is not a phoneme in the other language.
But a speaker of one language who is past the age of puberty will simply not perceive many of the phonemic distinctions in sounds in the target language (the language to be learned) without very careful training, as disregard of those distinctions below the level of conscious attention is part of having the sound system of the speaker's native language fully in mind. Attention to target language phonemes has to be developed through pains-taking practice.
It is brutally hard for most people (after the age of puberty, and perhaps especially for males) to learn to attend to sound distinctions that don't exist in the learner's native language. That is especially hard when the sound distinction signifies a grammatical distinction that also doesn't exist in the learner's native language. For example, the distinction between "I speak" and "he speaks" in English involves a consonant cluster at the end of a syllable, and no such consonant clusters exist in the Mandarin sound system at all. Worse than that, no such grammatical distinction as "first person singular" and "third person singular" for inflecting verbs exists in Mandarin, so it is remarkably difficult for Mandarin-speaking learners of English to learn to distinguish "speaks" from "speak" and to say "he speaks Chinese" rather than * "he speak Chinese" (not a grammatical phrase in spoken English).
Most software materials for learning foreign languages could be much improved simply by including a complete chart of the sound system of the target language (in the dialect form being taught in the software materials) with explicit description of sounds in the terminology of articulatory phonetics
with full use of notation from the International Phonetic Alphabet.
Good language-learning materials always include a lot of focused drills on sound distinctions (contrasting minimal pairs in the language) in the target language, and no software program for language learning should be without those. It is still an art of software writing to try to automate listening to a learner's pronunciation for appropriate feedback on accuracy of pronunciation. That is not an easy problem.
After phonology, another huge task for any language learner is acquiring vocabulary, and this is the task on which most language-learning materials are most focused. But often the focus on vocabulary is not very thoughtful.
The classic software approach to helping vocabulary acquisition is essentially to automate flipping flash cards. But flash cards have ALWAYS been overrated for vocabulary acquisition. Words don't match one-to-one between languages, not even between closely cognate languages. The map is not the territory, and every language on earth divides the world of lived experience into a different set of words, with different boundaries between words of similar meaning.
The royal road to learning vocabulary in a target language is massive exposure to actual texts (dialogs, stories, songs, personal letters, articles, etc.) written or spoken by native speakers of the language. I'll quote a master language teacher here, the late John DeFrancis. A few years ago, I reread the section "Suggestions for Study" in the front matter of John DeFrancis's book Beginning Chinese Reader, Part I, which I first used to learn Chinese back in 1975. In that section of that book, I found this passage, "Fluency in reading can only be achieved by extensive practice on all the interrelated aspects of the reading process. To accomplish this we must READ, READ, READ" (capitalization as in original). In other words, vocabulary can only be well acquired in context (an argument he develops in detail with regard to Chinese in the writing I have just cited) and the context must be a genuine context produced by native speakers of the language.
I have been giving free advice on language learning since the 1990s on my personal website,
and the one advice I can give every language learner reading this thread is to take advantage of radio broadcasting in your target language. Spoken-word broadcasting (here I'm especially focusing on radio rather than on TV) gives you an opportunity to listen and to hear words used in context. In the 1970s, I used to have to use an expensive short-wave radio to pick up Chinese-language radio programs in North America. Now we who have Internet access can gain endless listening opportunities from Internet radio stations in dozens of unlikely languages. Listen early and listen often while learning a language. That will help with phonology (as above) and it will help crucially with vocabulary.
The third big task of a language learner is learning grammar and syntax, which is often woefully neglected in software language-learning materials. Every language has hundreds of tacit grammar rules, many of which are not known explicitly even to native speakers, but which reveal a language-learner as a foreigner when the rules are broken. The foreign language-learner needs to understand grammar not just to produce speech or writing that is less jarring and foreign to native speakers, but also to better understand what native speakers are speaking or writing. Any widely spoken modern language has thick books reporting the grammatical rules of the language,
and it is well worth your while to study books like that both about your native language(s) and about any language you are studying.
 To refresh everyone's memory, pg's essay is about what happens when founders have trouble making themselves understood, while tokenadult's comment is about how to learn a foreign language.
As far as I know, it remains speculation that this effect occurs in humans, but it seems by your experiences (and my own learning Mandarin as a second language) plausible that the human brain itself "prunes" less useful sonic motifs in early life improving your ability to rapidly understand your primary language(s) even under shifts in accent or other distortions. This would likely come at the cost of literal inability to be attentive to motifs that haven't been stored. you would perceive them in A1, but they would have less resolution and perhaps semantic meaning (this is as far as I know quite wild speculation, mind).
I remember that it took almost an entire 6 mos of study before I really had a clue what tonality in Mandarin meant—I could hear it, but never perceive it as a linguistic phenomenon. After 6 months it appeared to almost overnight become something sensible to me at which point I had to relearn almost my entire vocabulary including the tonal information that I'd been unknowingly ignoring.
I went to Beijing for college, started speaking Mandarin, and picked it up quickly. But until now, I still regularly make the mistake between /z/ and /zh/, and the other two pair. I am able to easily tell the differences between the sounds themselves. If the sounds are included in a sentence, however, I simply can't tell which is which.
Here is the interesting part. I have no difficulty in distinguishing these sounds in English. I can clearly hear the differences between words like 'sip' or 'ship', no matter if they are spoken as single words or are part of a sentence. My ears will immediately catch the difference. But if it is Mandarin, i will get lost between words like 'ziji' and 'zhiji'.
As tokenadult mentions, it is brutally hard to learn these distinctions. perhaps that is why successful founders push their way past accent problems; they are successful as pushing through brutally hard problems.
When pronouncing an 'l' (alveolar lateral), my tongue stays on alveolar ridge of my mouth (the roof, right behind the teeth) for the entire time. When the 'l' sound ends, my tongue leaves the roof of my mouth, but it never "strikes the bottom" of my mouth. If there's a vowel next, it moves to a neutral position for a vowel. If the 'l' was the end of the speech, it just kind of sits in place for a second after the noise stops. If it's followed by another consonant, it moves for that articulation (e.g. when saying "all the sounds", it moves directly from the alveolar ridge to the teeth for the dental fricative).
When pronouncing an 'r' (alveolar approximant), I, and I think most English speakers, roll the tongue backwards to some degree (although not enough that the point of articulation is the bottom of the tongue). I.e. the tongue does not "stay at the bottom". Yes, it is certainly possible to pronounce an English 'r' with the tip of your tongue held down (unlike 'l'), but I think most English speakers would find this "impossible" at first, and then with a minute of practice would realize that it was possible but difficult. In English, these two articulations are perceived as the same phoneme; I assume that some language somewhere differentiates between them. I notice that often when I speak (English) to natives of India (whose first languages I have not identified), I get the sense that they have a richer complex of 'r' noises than I do.
In summary, "l" does not involve "striking the bottom [of the mouth]", and "r" typically does not involve the tongue staying at the bottom of the mouth, although it can.
Is that more clear? I agree that there are other r and l sounds in English, but there aren't in Japanese, there is a fixed set of phonemes. I'm trying to explain how each Japanese l/r sound is clearly split into a different l and r sound in English. Of course, there are many English l and r sounds that are badly approximated by the five l/r sounds in Japanese.
Anyway, my point was this: your description of English articulation had some minor errors in it, which I tried to improve upon.
> If I pronounce la (laa) ... in English, my tongue ... at some point touches my gums behind my bottom teeth
That may be true for you, but this is not typical. Or at least, it's not the the "L" that's doing that. If you say "raaaaaaaaa" in English, you'll find that the "aaaaa" noise puts your tongue on the bottom of your mouth just as much as it does for "laaaaaaa".
And if you put your finger on your lower gums, you can learn to say "la lee loo lay low" without your tongue ever touching your finger, much less your gums. Just like how many English speakers curl their tongue to say English "ra ree roo ray roe" but they can learn to leave the tip of their tongue down (with practice).
I'm pretty sure I know what my tongue is doing in English, but alright it may be due to the vowel instead of the consonant. The problem in Japanese is there's no distinction between vowel and consonant.
Yes, it could be that touching the bottom of the mouth during the vowel part is not as important as the position of the tongue initially. l: relaxed, some sliding, top of mouth; l/r: tighter, as for a rolled r, more forward, top of mouth; r: does not touch top of mouth. loo and roo are both incorrect pronunciations of ru.
The more complete description of the Japanese r is that it is a tapped alveolar consonant that can range between a completely central [ɾ] and the lateralized [ɺ], with different degrees of lateralization. All will be interpreted as the Japanese r, and speakers tend to use [ɾ] after vowels, with lateralization likely to creep in if the r comes at the beginning of an utterance or after an n.
I'm a native speaker of Korean by the way, and while at the abstract phonemic level we also have a single l/r sound represented as ㄹ in the Korean alphabet, we do distinguish [ɾ] and [l] between vowels. 아리 ari uses the tap [ɾ] and 알리 alli uses [l] because that's how double ㄹ is pronounced. It's a similar distribution to the tapped and trilled r's in Spanish, where trilled r's between vowels can be analyzed as double r's.
Also, our "r" is much thicker than in English. For that reason, our English accent resembles a bit that of Russians :-)
[it's just that the phrase 'second auditory cortex', especially in the context of discussing a process of acquiring a second language sounds suggestive of a development of another separate part. and A2 is definitely not that.]
This paragraph in particular:
It is brutally hard for most people (after the age of puberty, and perhaps especially for males) to learn to attend to sound distinctions that don't exist in the learner's native language. That is especially hard when the sound distinction signifies a grammatical distinction that also doesn't exist in the learner's native language. For example, the distinction between "I speak" and "he speaks" in English involves a consonant cluster at the end of a syllable, and no such consonant clusters exist in the Mandarin sound system at all. Worse than that, no such grammatical distinction as "first person singular" and "third person singular" for inflecting verbs exists in Mandarin, so it is remarkably difficult for Mandarin-speaking learners of English to learn to distinguish "speaks" from "speak" and to say "he speaks Chinese" rather than "he speak Chinese" (not a grammatical phrase in spoken English).
One thing I've noticed is that there seem to be relatively few super-successful Chinese immigrant founders in the USA. Fewer than Indians for example, even though both are large immigrant communities. My wife is Chinese and her theory is that Chinese aren't really good at teamwork. I think it has more to do with their difficulties in catching the subtleties of the language as described above.
It is not quite a problem of them not communicating their message well, I think. Most of the time they do, although it may not aesthetically be very pleasing. It is more an issue of them missing alot when you speak to them, unless you conciously try to adapt to your audience.
I also have noted that South Asians have a strong culture that resists rote Westernization, which I attribute to native Indians (etc.) learning over time about what to sustain and what to adopt from the culture of their British overlords. So I conclude that when South Asians emigrate to America, they have a "leg up" in adopting to American culture, while East Asians experience culture shock to a much greater degree. Not only is the language more unfamiliar, but East Asians don't have the historical/cultural experience of keeping their culture separate from a Western culture as South Asians do. As an example, look at the relative interest of children of immigrants in the USA's National Spelling Bee.
I therefore think that because of these reasons, and because I think East Asian immigrants are as bright as South Asian immigrants, that we will observe a bolus of super-successful East Asian immigrants one generation later than when we observe a bolus of super-successful South Asian immigrants.
Now obviously these observation are generalizations, but I suspect that the "strong accent" observation of PG is not just a proxy for facility with English but a measure of cultural adaptation than may be helpful in becoming successful in the USA.
It is more an issue of them missing alot when you speak to them
Language grasp =/= communications skills.
(1) Plenty of studies on body language and other forms of cognitive bias back this up. Attractive people are deemed more trustworthy etc.
(2) There are huge swathes of social interactions and nuances that cultural signals. 'Westenized' children educated for example in the us or uk grasp intuitively things completely alien to their parents or to similar kids brought up in the east. Such examples of ""comprehension has noting to do with wether or no the kids have accents or their "language skills".
(3) A regional welsh, working class, or a northern uk dialect is almost indecipherable to many americans. But such would not likely be a signal that this person won't be able to pick up westernized social cues.
I started learning English when I moved to the U.S. three years ago but I speak better than those who spent 10 years learning English in my home country.
This is verifiably true, by an experimental test of Chinese-language proficiency, of the foreigners who learned Chinese in my generation. Harvard and some other elite universities wanted to develop a test of Chinese as a second language to find out which Americans were learning Chinese the best. During the norming study for the test, someone thought to include foreigners at the Mandarin Training Center of the National Taiwan Normal University ( 國立台灣師範大學國語教學中心 ) in the norming sample. My fellow students and I who were there at the time "wrecked the curve" for all the graduates of Chinese language programs at United States elite universities who had not spent significant time overseas. Further development of the proposed test was scrapped after that was discovered.
That said, English is more learnable in-country than most languages because of its extensive use as an interlanguage. I lost count early of the number of different native language pairings I would see among foreign students in Taiwan--mostly in Taiwan to learn Chinese--who would converse with one another in English, because English was their strongest language in common. English-language movies, books, and other authentic examples of use by native speakers are also pervasively available around the world in a degree unmatched by materials in any other language.
My point is, if it's not true for English, then it's not necessarily true for other languages either. You just have to be interested in the culture, and expose yourself to media during and after taking classes.
Also, many languages leave their mark on their native speakers in many cases being very hard to get rid of your native accent.
I was lucky to have Romanian as my native language, as it doesn't leave such a big scar on your pronunciation. I almost speak American English correctly, in spite of not living in an English-speaking country and I've got friends that speak perfect British English, French or Spanish (giving these as examples, as these have thick accents). True story - Microsoft has a support center in Bucharest, with one reason being our linguistic abilities.
To this day, I'm still baffled that Spanish does not distinguish between 'v' and 'b'.
Interestingly, my native language is Czech, and it makes it rather easy to learn English and Spanish, because the only sound that's missing is English 'th' in three or think, which people here pronounce like 't' or 'f' and "I fink" sounds pretty horrible :)
I understand that Spanish has to be really hard for English speakers, because of things like words changing shape because of gender, and stuff, but the usual English/American pronunciation of Spanish 'j' (or 'x') is terrible. I had to laugh at Lady Gaga singing about some Alexandro, making up about three different ways to pronounce it, not a single one correct. It's not difficult sound!
Talking about difficult sounds, try this one:  it's fun :)
You forgot people mixing up v/w, and being funny with r. ;-)
My only imperfections that I and my GF (who is from the UK) recognize are things like my US-centric pronounciation and vocabulary ("boot, not trunk!", "lift, not elevator!", etc.) and my tendency to mispronounce words that I've only ever read but not heard. And when I'm tired, my accent sometimes slips and I suddenly stereotypically sound like Arnold Schwarzenegger.
This is before they train on how to fly the Soyuz, and the Astronauts do serve as flight engineers on actual Soyuz missions to the ISS. There is a reason NASA does this.
Plus, if you live in NYC or any ethnically diverse community in the USA, you'd meet plenty of people who've lived here 20+ years, yet can't have a conversation with you in English - I know many Chinese immigrants like that.
First step for learning a language is wanting to learn I guess
I wish I had internet and wikipedia when I was in school.
Patently false, as evidenced by the countless numbers of second language English speakers who have never been to an English-speaking country for more than some vacations (if even that).
he chose the word "community" precisely to indicate that you don't have to live in an English-speaking country to learn good English (or a Chinese-speaking COUNTRY to learn Chinese, etc.). But massive "real-world" exposure to a language outside required school lessons helps immensely, as many research studies on language acquisition around the world have shown.
English has the network advantage of being the "community" language of people who have no other common language, all over the world.
Maybe the primary take-away from those findings should be that learning languages in school sucks. I can still remember being assigned 20 words to mindlessly repeat until I memorized them.
There is a simple and effective technique to help adult learners to perceive difficult phonemic distinctions, called HVPT (High Variability Phonetic Training). It consists of listening to several native speakers produce the phonemes in question and being quizzed on which phoneme is used in each case. You receive immediate feedback whether you were correct or not. This turns out to be much more effective than simply listening to a single speaker pronounce the phoneme pair and being asked to hear the difference (multiple speakers are crucial in helping you generalize what the salient differences are for a range of pronunciations).
Unfortunately, I don't know of any actual implementation of HVPT, and it doesn't seem to be used much if at all in language learning. Surely there is an opportunity there for someone to design a web-based HVPT system.
I'm now working on creating something better. In the meanwhile, getting a book with good phonetic descriptions of your target language (including lip/tongue diagrams) and recordings, and doing phonetic transcriptions is a better bet than HVPT. Being able to correctly produce the sounds goes a long way towards being able to distinguish them, though it seems to be a two-way feedback loop.
This is I feel a relatively neglected part of non-native language education, and it is very common for people to have spoken a non-native language for years without learning to distinguish native sound pairs. So good luck with your efforts to help people learn better, and I hope you share the results with us.
You could also say that learning to use a computer is key to become a programmer. That's obvious and not very telling about what is specifically required to be a programmer.