Hacker News new | past | comments | ask | show | jobs | submit login
Best ways to learn a foreign language (domenicodefelice.blogspot.it)
74 points by domedefelice on June 30, 2013 | hide | past | web | favorite | 35 comments

Professional translators often do "shadowing", which is listening to source material from the language you're weaker at and attempting to just transcribe it in real time. You can start with source material which is easy, such as dialog on TV shows, and move to source material which is hard, such as news broadcasts. (Newscasters speak much much quicker than characters in fiction and employ large and often unpredictable vocabularies, whereas most popular culture is dominated by the ~200 words at the head of the zipf distribution for that language.)

It's totally free and, when you're at intermediate proficiency, both stupendously effective and the most maddening exercise you'll ever do. (What happens when you're thrown off the deep end into a newscast about nuclear power plant issues and you're deeply out of your depth with regards to the required vocabulary? You try your best, jot down the five words every 3 sentences you actually catch, and then break out a dictionary later to figure out why houshasen and shiyouzuminenryoupuuru seem to be so key to the topic.)

("Radiation" and "pool (for storage of) used (nuclear) fuel", respectively. And there, now you understand 5% more of that conversation.)

A small thing I've noticed about seeing news from France and Switzerland.

Swiss newsreaders seem to speak more slowly. I don't know if it's a peculiar Swiss behaviour, but I wondered at the time if it was connected to the fact that Switzerland has four official languages, some of them having overlapping linguistic footprints.

To make a Swiss French broadcast accessible to Swiss Germans who have rudimentary French, it would make sense for the presenters to speak more slowly.

Meanwhile, over on France 2, the appallingly videogenic presenters of Le 20 Heures drive a swishy gallic stereotype through the paper thin limits of my understanding.

The French have this stereotype of the Swiss being slow-spoken (some go as far as to claim they are slow-witted). So it's not just newscasters, but the general public as well.

Similar to dictation but seems funnier and more challenging at the same time. Very interesting! I am going to try this

Learning a language isn't just a process of learning words out of context (with spaced repetition systems like Anki, or products that piggy back off of it like Duolingo). I don't mean this in an unkind way, but I get the impression the author hasn't ever learned a foreign language very well as an adult.

Until I became an engineer recently, my entire career revolved around learning and teaching foreign languages. I strongly recommend looking at what L2 acquisition linguists have learned before hitting bloggers for language learning advice. One great resource is Dr Krashen's website. The video on the front page does a great job of showing how context and comprehensibility matter more than brute memorization algorithms: http://www.sk.com.br/sk-krash.html

And if you must learn from a blog, pick a blog of someone who has learned a lot of languages well (e.g. Steve Kaufmann) over one who talks of his "learning hacks" but doesn't any videos of himself speaking a foreign language in an unscripted setting. That's my 2 cents.

Anki (or any other SRS) doesn't have to be about learning words out of context. I've learned a lot with Anki plus sentences off of Tatoeba, which has a large amount of sentences translated into many languages. And I did that while being familiar with the basics of Krashen's ideas. I think there's no contradiction. Learning words and not forgetting them helps me to focus on getting a lot of input on my target language, without the drudge of (re-)looking up the dictionary for every other word.


By the way, (for the benefit of other people reading this) the website you link to has many other interesting articles about language learning vs. acquisition.

I completely agree with this.

I've created the site Readlang (http://readlang.com) to allow learning vocabulary in context by reading uploaded content, translating, and using SRS flashcards which include the original context.

It's in public beta and would love to get feedback if anyone tries it out.

I checked out your site about the time it was announced on Reddit. I really like the idea although I noticed some problems using it for my target languages (Spanish and Hebrew). For inflected languages like these, you really need a good stemmer/lemmatiser (either one, I'm not an expert, but you get the idea). I haven't tried Readlang recently though, when I do I'll be happy to give you more feedback.

Thanks, yes I'm relying on Google Tranlsate at the moment, but eventually I'd like it to recognize word families and verb conjugations. I use it for Spanish and regularly edit the generated verb flashcards to indicate the correct conjugation.

I've used Memrise and Duolingo to learn a bit of a couple different languages, and I wouldn't really say they're out of context. Even though memrise is a flash card app at heart, it gives you words and full sentences using those words at the same time. Duolingo does the same, and goes as far as showing you how to conjugate those words.

Hi Xiaoma, thanks for your feedback. I think that Duolingo and Anki are great trampolines to learn the basics of a language. Duolingo worked great for various people I know and it helped me learning the basics of German in a funny way.

I've made good use of Anki myself. In fact, I was an early contributor to it!

I still feel SRS is best used as a supplement instead of primary source, though. In Taiwan I knew a guy who had spent 3 hours a day for a year and a half memorizing Chinese characters and phrases on an SRS. He could write 3000 characters. He wasn't that great in terms of actually communicating with people or reading comprehension, though.

This came up in HN thread on a language teaching app about 6 months ago: https://news.ycombinator.com/item?id=4992561

Let's assume he took weekends off, so he had 3 x (365 + 183 - 150) = 1194 hours of study. In that time he learned to read and write 3000 characters in context.

How is this a bad example of language aqcuisition? It sounds to me like the guy learned what he was studying, and learned it very efficiently.

If I remember correctly, most school curricular will have you learn 3000 characters in 3 years, and most students won't be proficient in them when finishing.

Ofcourse, if his goal was to chit-chat about the weather, then obviously he wouldn't learn that since he didn't practice it, but it sounds like he could read a newspaper after only 1.5 years of study, which I for one find rather impressive.

>In that time he learned to read and write 3000 characters in context.

Well actually he didn't learn them in context! That's the thing. He absolutely could not read a newspaper if by read you mean understand the content of the articles. He could mostly sound out sentences, with poor pronunciation and often using the wrong reading for various characters (e.g. pronouncing 行 as háng when it should be xíng or vice versa).

He could write many, many characters based on English keywords and pinyin associated with them in his deck. But as I've been trying to explain, words don't map one to one across languages and learning a language isn't simply a matter of memorizing words. The case of using a deck for isolated Chinese characters is even worse since a character is a syllable as opposed to a word. In modern Chinese, the vast vast majority of words are poly syllabic (i.e. multi-character compounds).

FYI, he had been studying some Chinese for a couple of years before going on the SRS craze. And he didn't take weekends off. He put even more time into study, fortunately including some things other than decontextualized SRS-based study. During the time I knew him, he largely backed off the SRS in favor of getting loads of comprehensible input in the form of media, talking to parents of his English students and his girlfriend's friends. Over time the input did lead to a better ear for the language, better pronunciation and the ability to read and actually understand. It's possible the first couple of years with the SRS gave him a good foundation that helped for the next three years of working with the language in context. I still think it would have been a better bet to focus on listening and reading from year one.

Based on your experience, what level of proficiency would you expect from 1.5 years of 3hrs/day study?

3000 actually tracks about right from what I know. Here's some information with little backing: I've read somewhere that children pick up around 10 new words a day, and at the peak of my full time language studies, that was what I was achieving as well. Considering that it takes a few months to ramp up to that speed (I've read that it takes about 10 separate exposures to a word for it to become memorable), his 3000 characters for 1.5 years is pretty good.

"Acquisition requires meaningful interaction in the target language - natural communication - in which speakers are concerned not with the form of their utterances but with the messages they are conveying and understanding." Stephen Krashen word. i made several failed attempts at learning a second language before up and moving abroad. after a year, i'm still pretty bad at it, but i don't know if i'd /ever/ have made it this far if i didn't have to speak on a daily basis... although part of that assessment is based on me being lazy and bad with natural languages.

anki and similar are really useful, though, if you're studying a non-phonetic writing system.

There is plenty of anecdotal evidence of what works for some individuals.

However, pointing out the common problems, shared by everybody, is also informative. Here are some:

* The factors in learning a new language are age (adolescence is a threshold), innate skill, motivation (and guilt), effort and similarity of the languages.

* Sounds are not listener-independent: your Chinese friend does not hear the same sounds as you do, so don't expect that they know which sounds you are correcting in their English when you repeat the word for the tenth time. Same the other way around. I have seen this leading again, again and again to absurd "are you deaf?" situations between otherwise attentive friends and teachers.

* It might be helpful to discuss metalanguage: direct objects, word order, verb transitivity, nouns vs verbs, parsing, etc. This is something you learn in school in France, Spain, Germany, some Arab countries, etc, and if you studied Latin. But if you went through a standard English-speaking school system you might want to catch up with that.

>don't expect that they know which sounds you are correcting in their English when you repeat the word for the tenth time.

This phenomenon is probably the most interesting thing I've found out in my life. The fact that some people simply cannot hear certain sounds goes against everything we would like to think.

When people talk about varying tone inside words (such as in Chinese or Thai), while I obviously hear that something is happening, I cannot for the life of me reproduce what I hear or even retain the tonal information. For me , tone is a separate aspect to vocabulary and I can't assimilate the two.

Inversely, seeing the plight of poor Japanese people, who get taught English from an early age through the limited japanese syllable system (not sure the exact name), unable to repeat words properly, and sometimes going completely off the mark.

This just accents the need for everyone to be exposed to as many languages as possible as a child.

I would disagree with much of this article. While Duolingo in particular is popular, its main focus is on teaching translation. It's hip, but it's not about teaching how to speak and understand a language. Similarly, the focus on vocabulary study in the other tools has similar limitations.

Self study needs to be augmented or driven by spoken conversation practice in the least. Your brain is a neural network to learn language as audio input. While challenging, it's the most effective way. Google doesn't support this due to the link popularity of a lot of software-driven methods.

Disclaimer: My business in the online language learning space.

+1 to watching TV with subtitles, and even better repeating out loud the sentences you find more interesting. That way y * It will be easier to remember and * You will improve your pronunciation

I have a language learning site[1] which started out as text translation, but I'm also experimenting with videos, the idea being it's helpful to see captions in both languages simultaneously as the video plays.

Examples here[2], here[3], and here[4].

[1] http://www.macaronics.com/

[2] http://www.macaronics.com/article/6e9812608f0b4dcdba8cfab3e6...

[3] http://www.macaronics.com/article/9540f24d080241dd9b27f7939c...

[4] http://www.macaronics.com/article/a6ab2c3d8c364ad38401d6dba3...

If in Asia, that means Karaoke!

you need a repetitive stream of natural language. then brain automatically analyzes it.

usually, you pick one 30 minute to 60 minute video, and watch it every day, 6 days a week. make sure you take a break every week. you watch it to the point where you can mimic what they are saying as the video is playing. you don't have to know exact meaning of your utterance. but you already have some sense. then, you transcribe it on paper (you need to learn writing system by this time). maybe use dictionary to learn actual words. rehearse with the video. if have friends, rehearse with them. memorize it. move on to the next video.

start with sitcoms then maybe news broadcast, reality shows (court shows are great), movies...

video is just one example. if you're learning a language where you can't find suitable videos (there might be no production in the language), make friends who speak the language. record their conversations. play repetitively until you can follow along with a loud voice. transcribe.. etc.

Didn't see any references for us Classicists so I'll add my two cents:

I'm still struggling with classical Latin after 8 years. The problem being that it's a "dead" language and immersion/conversation is difficult to achieve on a modern era time schedule.

Over this time I've come to the conclusion that you need to start "speaking" in any language as soon as possible. The approach of learning grammar for a year and translating sentences on pen-and-paper is not effective enough.

My current toolbox for learning Latin is:

1. Supermemo (Anki is a free equivalent) for vocabulary and anything that needs to be committed to memory permanently. This takes away the burden of worrying about anything that needs to be remembered.

2. Follow Evan der Millner's "Comenius" Latin project, which attempts to teach you Latin the way school children were taught back when they were expected to speak, read and write it fluently. This is a reconstruction of the way Comenius would have taught children. [1] project overview, [2] reading of the texts. [3] Oral lessons, so you can start Skyping other Latin speakers and talking as soon as possible. (this is not "church" or Italian pronunciation, but the received Classical reconstructed pronunciation). Everything you need is available for free off the either Google books, Evan's YouTube videos and archive.org.

3. Know and use Diogenes (http://www.dur.ac.uk/p.j.heslin/Software/Diogenes/) for Windows or Linux.

4. Install the "Language Immersion for Chrome" extension and set it to Latin so you're forced to read the Internet in Latin as you browse ;)

[1] http://latinum.weebly.com/comenius-project.html [2] http://www.youtube.com/playlist?list=PLMH5SfME31ZKYChsIWSp_D... [3] http://www.youtube.com/watch?v=y0NIMm2eM8c&list=SPC6E7F1C5D4...

I'm attempting Attic Greek this way (Evans has started an Oral ancient Greek course on YouTube too) and can't wait to enjoy reading Plato, the Iliad and the Odyssey aloud in the reconstructed native tongue in my old age some day :)

I have downloaded and tried to play with Diogenes several times. I can never seem to get it to do anything. As best as I can tell, you need to provide it with data sets as well. Do you know where you can get these?

You don't need the data sets unless you're interested in exploring the Greek or Latin texts. On it's own it's useful for entering dictionary words in their declined/conjugated forms and it finding the correct entry(ies) in the dictionary.

For the data sets: In all honesty, just torrent it. What you're looking for is TLG /PHI/ Cd-rom_E, which is the Thesaurus Linguae Gracae and the Latin texts.

Wow, you're right. It does come with some data out of the box after all. Its UI isn't all that it could be, though.

> One of the best ways to improve your expressing skills is to talk with natives.

Go to the foreign country and don't speak one sentence in your native language. Jump in at the deep end (BTW, I had to look up that phrase because I never lived in an English-speaking county).

Having a general interest in languages, I've tried many things to improve my language learning. The thing I was probably most impressed with though is the Michael Thomas method.

The submitted article had some interesting tips, and several of the previous comments are quite good too. I've been developing a FAQ on language learning as this interest is mentioned on Hacker News from time to time. 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 that 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.

Getting an ear for a foreign language is quite difficult - getting to the point where you can separate the words out (without necessarily knowing what they are) so that you can learn them. I have quite good reading French (I can read proper full-length books) pretty poor written French and bloody awful spoken/hearing French. Listening to a lot of French pop music on Spotify really helped me get an ear.

Hack 0 should be to find a good motivator to learn the language. Only then do tools matter.

I got bored learning Kanji. Had I known that 300 of those were enough to read a food menu -- and what they were, I would have prioritized them and learned them in a month. Instead, I stopped after thinking it would take too much time to ever get to 2,000.

I'm quite happy with rosetta stone. do you have any particular reason not to like it? (beside paying?)

Hi bigd. I've only tried the free demo of rosetta stone and it was too short to make me form a useful opinion about it. Anyway I would prefer free material.

I find that using mnemonics is an efficient way of learning words. The only problem is to find good mnemonics, which can be easy if the word sounds like something you already know, and hard if it reminds you of nothing and the word represents a very abstract concept.

Memorizing something like the top 1000 words used in everyday conversation, with little grammar exercises sprinkled in, seems to be a good strategy. I've only got to memorizing something like 500 words (give or take 150 words...) so I haven't put this method quite to the test yet.

Applications are open for YC Summer 2019

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact