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 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
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.
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.
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.