The Immersion Delusion

Depends on your circumstances. I have three, hour-long Italki lessons a week. I speak to Chinese friends every day and they are happy to correct me when I make mistakes. I have been fortunate to meet many patient friends who have helped and continue to help me immensely.

Although maybe it is just the insane difficulty of Chinese…maybe I do need to read 50k pages (my plan) and listen to 10,000 hours of audio over many years to be able to speak effortlessly?

No, it’s not that. All the evidence suggests your instinct is correct. You can’t speak because you haven’t practiced. This is a particularly an issue for Chinese where failure to become comfortable with the tones leads to distortions in the way we percieve and produce them.

This can be overcome, of course; In language learning no damage is irreperable. You can retrain your brain to percieve and produce tones correctly. But it’s a bad idea to believe that more input alone will solve the problem as this doesn’t seem to be the case.

For more on how I overcame similar issues check out my blog here: It’s Never Too Late to Learn Chinese Tones. Here’s How – I'm Learning Mandarin

I’ve been thinking about this, and I really think it comes down to personality and goals.

The mass immersion approach has been embraced by certain parts of the Japanese learning community, because it gives them permission to do what they are learning the language for - watch a lot of anime and read a lot of Japanese books. It was a revolution because a lot of the traditional approaches involved a lot of talking to people, moving to another country or learning from resources that don’t have relevant vocabulary. Speaking is like a cool party trick / side benefit, not the end goal.

I don’t see a lot of examples of people doing this for Chinese (except the Heavenly Path site) since a lot of people seem to be learning Chinese for career purposes or the intellectual stimulation itself, not because they want to read danmei webnovels or are mad at the quality of Mango TV subs.

This conversation has been super helpful, because now I realize my crappy listening skills are a problem… I shall now turn on pinyin and do that listen for an hour to a podcast with transcript at 1.3 speed that the Peter (?) member swears by.

I really think it comes down to personality and goals.

  • You’ve hit the nail on the head. Extreme immersion started out as a reaction to traditional dogmatic views on language learning being imposed on everyone regardless of their personality type. Using traditional methods, people who don’t like speaking or studying grammar don’t get to learn languages. Extreme immersion offered these people another way. But the movement morphed into something equally dogmatic whose purpose became to convince the world extreme immersion and extreme immersion is the only way to learn a language. This idea has mass appeal because of how much people hate traditional methods. People didn’t just want alternative methods that could help them learn, they wanted a campaign to wage war on traditional methods. Anyone who questioned these ideas or supported methods became the enemy.

I don’t see a lot of examples of people doing this for Chinese

There is another reason you don’t see people doing this for chinese. With Japanese it’s possible to have bad pitch accent - which most people who follow extreme immersion methods do - and still sound phenomenal. But the equivalent person learning Chinese would speak with poor tones and not only sound atrocious, but not even be comprehensible. For this reason there are no role models in the Chinese learning community to follow. People always look to follow the methods that the best speakers use.

For a while some believed the success of the All Japanese All the Time AJAAT movement would lead to a spin-off ACAAT (All Chinese…) version. But it never happened. Why? One answer is tones. Japanese had pitch accent but you can still reach an excellent level and be understood without it. Without Chinese tones, however, you can’t even get off the ground. There may also differences in terms of the extent to which the grammatical patterns of both languages can be acquired through immersion alone. Either way, I have yet to see AJAAT results replicated for Chinese or, for that matter, european languages.

Hi TofuMeow,

“I shall now turn on pinyin and do that listen for an hour to a podcast with transcript at 1.3 speed that the Peter (?) member swears by.”

We’ve enough evidence so far that (what we call) the “ultrareading while listening” (URL) approach (with an audio speed > 1.0x) works pretty well for not so distant Indo-European languages. But we need more experiences / stats with distant L2s (Mandarin / Cantonese, Japanese, Vietnamese, Arabic, etc.).

So it would be great if you could share your URL experiences (and maybe some stats) on the LingQ forum in a few weeks/months!

“to see AJAAT results replicated for Chinese or, for that matter, european languages.”
Before AJATT, there was “Antimoon” (for English, etc.):

However, Antimoon also mixes both artificial and natural SRSes…


I think you speak Spanish as a bilingual kid if my memory serves me right. You can evaluate his level in Spanish.

Fluent in Spanish in 10 Months | Matt vs Japan Interviews #9 - Khalifa

“the immersion delusion” is a bit more complicated, IMO:

  1. “Early outputting”, esp. in schools, often fails to this day (!) because students simply don’t have enough exposure to their L2. Therefore, they tend to translate from their L1 into their L2 and form sentences that are semantically “off”.

To be more precise, these students tend to create word combinations that often sound unnatural because they don’t know the highly conventionalized (“formulaic”) word groups (esp. “collocations” and, to a lesser extent, “idioms”) that native speakers use when they communicate.

  1. The solution to this problem were input-heavy approaches (based esp. on Krashen) such as Antimoon, AJATT, Mass Immersion Approach (MIA), or Refold.

Main advantage:
Adopting an input-heavy approach will definitely improve the learners’ non-conscious knowledge of collocations. So when they have been exposed to the language for thousands of hours, they will have a good sense of how correct sentences should sound in their L2.

Some problems with such approaches:

  • a high risk of burning out.
  • the danger of low-quality listening, i.e. listening without much attention (which is rather a waste of time) [see our discussion about “passive” listening on LingQ from a few months ago].
  • the conclusion that any early output is “bad”, which is nonsense, because the specific problem was “early output without sufficient L2 exposure.” However, if learners couple early outputting with enough input and / or with a feedback loop based on deliberate practice à la Will Hart (i.e. saying something → getting corrections → repeating the corrected sentence(s) and / or using an explicit SRS for learning these corrections → rinse and repeat) etc. then “early output” doesn’t do any harm. On the contrary, it improves the level of engagement with the L2 compared to relying exclusively on reading and listening activities!
  • when heavy input adepts postulate that L2 learning requires a “silent period” similar to that of babies. However, human babies are born too early, so they must mature outside their mother’s womb. This isn’t the case for teenagers or adults. Therefore, for the latter, a silent period of many months or even years is unnecessary. Instead, they can speak (and sometimes even write) from day 1 of their L2 learning journey while avoiding the pitfall of “early output without sufficient L2 exposure” (see the point mentioned before).
  1. The real questions in this context aren’t related to “magical thinking” like
  • becoming orally fluent by inputting alone and without any speaking practice
  • (similarly): becoming a good writer by inputting alone and never practicing how to write (well).
    Reading, listening, speaking and writing are all “practical skills”. So if learners don’t practice them, they can’t become good at them (however, there are some spillover effects, for example from listening / reading a lot to speaking and writing!).

The real questions are rather:

  • Should we output early or later? (and that’s not a question of effectiveness - because both approaches work! -, but of efficiency!)
  • Which is the right input - output mix depending on the personality of the learner (shyness, etc.) and the language level (beginner - IM - ADV)?
  • Should we primarily rely on unconsciousness learning (see, for example, Steve’s latest video on how he learns vocabulary based on Krashen’s research)?
  • Or should we also learn consciously by using artificial SRS, etc., e.g. for non-frequent collocations, important grammar patterns, etc.?
    My personal take based on my learning and teaching experience is:
  • output as early as possible, but combine it with a lot of input. esp. as “(ultra)reading while listening based on content flexible audio readers (à la LingQ, ReadLang, etc.)”.
  • combine unconsciousness learning (reading / listening as “natural SRS”) with conscious learning (i.e. artificial SRSes à la Anki, Memrise, etc. for the most frequent 1000-5000 words / sentences at the beginning, explicit pronunciation activities à la Idahosa Ness, and, if necessary, “grammar light” approaches à la Michel Thomas, etc.).
    So I’m definitely in the “(flexible) hybridization” camp when it comes to learning in general and language learning in particular :slight_smile:

Nice WE to you all

There the author writes about the “language module” in our brain, just as we have the “face recognition module”.
But then he jumps to an implicit (if not explicit) conclusion, that it all boils down to an amount of input.

But it’s not that simple. We don’t have just one general “language module” across the board. We have different and highly specified neural populations, that responsive, for example, to singing, but not instrumental music or speech etc.
(LingQ interrupts with URL of this link. Pls, just copy and past, don’t click it)
(“the link you absolutely have to see before you’re ready to understand my point” :D)

Input would be like a messy database, but not a substitution for the actual skill, when it comes to speaking.
It’s like in chess. You’ll reach a higher level of skill just playing chess and rarely watching tournaments or reading theory comparing to those who’s mostly watching tournaments and reading theory, but almost never played it themselves.

The reason, why we seek for some kind of substitution, is what @asad said. “It is extremely expensive to hire a tutor every day”, and sometimes it just isn’t realitic for some of us to do what Will have done.

For example, I rather can learn Chinese that way, than English. Because language Exchange “Russian-English” just doesn’t work, not so many English natives are interested in learning Russian beyond “Privet kak dela”, 95 (actually 100%) are males, interested in dating, not the language exchange.

Chinese people learn Russian for business reasons, at least.

Hiring English tutors for daily intensive practice is expensive, because of the difference of value between Dollar and Ruble. It would be what, like $4.500+ a year? Sure it should be effective for language learning :smiley: It better be!

Some people might have other reasons. It just that there no one size fits all approach.

Our life is a complex thing. All those claims of some young enthusiasts, that they have finally found THE METHOD are at best just marketing for their content on social media, at worst it’s just short-sightedness and obsession.

Input is good. Output is good. Vodka, Beer, Bourbon all are good. Girls, boys, cars, airplanes and birds all are good. Searching for what is good FOR EVERYONE usually isn’t good. History knows many figures, who knew what is best for everyone. It all starts like this. Today you don’t wanna learn languages with input/output, tomorrow you’re in the GULAnG somewhere in Siberia.

“Input is good. Output is good. Vodka, Beer, Bourbon all are good. etc.”
Just mix everything together (input, output, vodka, beer, bourbon, girls, cars, etc.) like I do then you never have to worry about “THE METHOD” :slight_smile:

BTW, the link doesn’t work [is that some secret Russian research for creating the ultimate Soviet brain - sorry, comrade: I couldn’t resist :slight_smile: - that we in the West aren’t allowed to see?]

The page or action you requested has resulted in an error. Please go back to the previous page by using your browser’s Back button, or visit the Home Page.

Here the link:

OMG, I can’t even delete this comment, nor can I really fix the link. Zoraaaan!

LingQ does something to the link. Try to literally copy/paste that link, don’t click

Just to chime in here again,
I can heartily recommend using podcasts to improve listening comprehension. Starting in April I have implemented LR-ing of podcasts into my routine. And, subjectively, I’d say my listening comprehension has improved quite a bit.
The inspiration for this, of course, goes back to Chaseb’s excellent post around that time.

I’ll share some of my experiences in the following, maybe someone finds them useful.

As for transcription services, Steve Kaufmann is known to use, I have not tried that, due to the low advertised accuracy (<85%) and the prohibitively high prices (18€ for 90 minutes). So who knows how good that service is, there are however many such transcripts available on LingQ Persian, maybe a Persian speaker could take a look?

Chase recommended but I have been unable to create an account there. But this seems to be well established in Mainland China.

The only service I have experience with is in fact Amazon Transcribe – Speech to Text - AWS and the results are actually pretty good. It fails, unsurprisingly, whenever multiple people speak at the same time, but it seems to handle basically everything else. Prices are reasonable as well, but that might depend on region and currency, for me it’s around 2,50€ for 90 minutes.

Next up on my list of services to test is, by Facebook is supposedly excellent as well, but I can’t bring myself to create an account there…

Actually, I wish LingQ, which uses AWS extensively already, would offer a seamless transcription service on the site, this would be supreme, but unlikely of course.

There is also a YouTube channel that showcases quite a few Chinese podcasts, there are even some videos, with accurate, subtitles. Those are however burned-in, so not usable on LingQ but I found them to be helpful nonetheless.

I have also shared some podcast episodes on the LingQ library in the past: Login - LingQ
Which I really like, the transcripts are so-so, but honestly anyone who is not a complete beginner should be able to handle that.
(Btw. I don’t think I will share long-form content on LingQ anymore, since the idiotic unfortunate 2000 words per lesson limit makes the whole process supremely annoying.)
Anyways, y’all have yourselves a nice weekend.

“Hiring {English] tutors for daily intensive practice is expensive,”
It doesn’t have to be. See:

Nice WE
Peter out

It’s a dating app

PS -
Since when is “” a dating app?

“With millions of members, Tandem is one of the largest language learning communities out there! With our Tandem app, you can connect with native speakers all over the world and practice languages via text, audio message, and video call. Download Tandem today - it’s free!!”

The thing I am interested in is really how it works with less familiar writing systems. Icelandic has a few letters that don’t exist in other Latin scripts, but I could follow along from day 1 even if I understood nothing. Not sure how Icelandic ranks for more distant L2’s in my case… (At some point in my life, I will find out how it works with Sami and/or Finnish.)

Is that possible with less familiar scripts at 1x early on? That I have no idea. My belief is that once that is possible, that the benefits are enormous.

By R+L until the advanced stage where I can simply listen, if nothing else, it lets me get exposure to more words per hour (12-18.000 vs 3-6.000 with TV) and guarantees I am focused and not thinking about who knows what.

It is just I have no idea when that becomes possible to start. Maybe a romanization above the text could be an idea to enable this even sooner?

Man, I answered this in the original comment. Free Language Exchange for Russian-English doesn’t work. I tried tandem and other exchange apps for half a year and yeild nothing. Just bored dudes waiting for Natashas and Lenas. Well, it works for them, but if I disguise myself as a Russian girl it won’t last long enough for fluency to be acquired, as long as I’ll start to speak. Also I’ll have to shave my beard. And it is something I can’t go for :smiley:

I tried HelloTalk and it felt like a dating app. I think the ideal would be find a good conversation partner/friend and then leave the app. Maybe even one where you have the same TL.