As promised, here is my full interview with Will Hart, a 20 year old medical student who on the eve of the first UK lockdown in 2020 had never been to a Chinese speaking country, had no Chinese family and had never had any meaningful contact with the language in any form.
Fast forward 12 months and he posted a short video to YouTube speaking with the kind of fluency many people fail to reach after more than a decade studying the language immersed in Chinese speaking countries.
How did he do it? Thatâs what I wanted to find out when I invited him on my podcast.
Are you afraid of early output and tempted by input purism? Wondering why youâre not fluent yet even though youâve read 100 novels? Frustrated that you can never remember corrections?
Whoever is gonna listen to the podcast, please describe it in a nutshell here in the comments. Here in Russia we have only the Big Brother Radio now, ainât no spotify-shmotify.
Itâs not an exaggeration to say it has made me rethink my entire approach.
I saw that the blog post which preceded this podcast generated a lot of discussion on the forum - I think that in itself demonstrates that there is something interesting going on here. As somebody who has been learning Mandarin for a while, I agree that what Will has done is pretty much unheard of. I am really excited about trying out some new methods off the back of this podcast.
One of the challenges in talking about language learning is that it can be hard to precisely quantify ability, and there are so many variables that each personâs experience will be slightly different, and can never be replicated exactly. So rich qualitative insight (like this interview) is often our best bet in giving a sense of what best practice might look like. Thank you for sharing it with us
Good conversation. Really appreciate it. A lot of good points to take away but for me, it was trying to interact with natives as much as possible and creating personal experiences. It helps you to think in your target language. Also doing those dull things like shadowing/imitating every day. Consistency is key.
Friends
Willâs circle of Chinese friends was not only a constant motivating factor but provided a source of meaning. It is easy to see language learning as an activity apart from ones ârealâ life, thus making it unnecessarily abstract. This reminds me of Benny Lewis, who deliberately put himself in high-stakes situations to speed up his learning. Iâm sure similar mechanisms are at play here.
Deliberately
By soliciting corrections from native speakers and making corrections to his speech, Will created a feedback loop. This can be very powerful, but how can one get Willâs result without access to a pool of infinitely patient, friendly natives? A potential solution might come from the YouTuber âLanguage Lordsâ, who created a self-directed feedback mechanism by recording himself speak, then watching the resulting video, noting mistakes, and then retaking a hopefully improved video. Maybe that can approximate the effect? This is at least what I plan to use to activate my Chinese.
Iâm glad Will mentioned the Shadowing technique, which used to be a favorite of mine. Since I have recently fallen out of the habit itâs good to get a reminder to pick it up again.
Chill
For a number of reasons, our hero, Will, has gotten really good really fast. These reasons may include his circumstances, his talent, his methods. Some of these components, we may be able to incorporate in our own study routines, some we may be able to substitute, some not. In the end I find it most important not to stress out and endlessly cogitate about the optimal approach to language learning or brood over other peopleâs successes. My choice is to just run my own race: less think, more do
Have a good one everyone.
Thanks for listening. Yes I agree completely. I think the best thing to do with these cases is to mine them for wisdom and then apply that wisdom to your own learning in a way that suits you best.
For me these are the main takeways from Willâs story I will be applying to my own learning in future and suggest for others who have a similar goal of learning to speak languages (especially Chinese) well::
1.) The âearly output is dangerousâ myth is dead and buried. Will alone didnât discredit it but the evidence is now so strong and there are so many cases that contradict it that I see him as the final nail in the coffin. Please nobody ever repeat this again. If you hear ever anybody state this myth as fact ask them for evidence and when they inevitably reply they donât have any ignore them and go on your way.
With Chinese, please for the love of god (if it isnât already too late) focus on tones and pronunciation from the start. Donât read novels before youâve learned how to pronounce properly. If it is already too late please read my in-depth guide for the majority of learners who have bad tones at intermediate-advanced levels. There are interventions that will help you get back on track. The road will be painful, long and arduous but these interventions are effective as I have found from personal experience: Itâs Never Too Late to Learn Chinese Tones. Hereâs How â I'm Learning Mandarin
Also with Chinese specifically, use a space repetition system that focuses on WHOLE SENTENCES and phrases rather than memorising thousands of words in isolation. Internalise the grammar in this way and DO NOT, under any circumstances, approach the challenge of Chinese grammar in a similar way to how you might approach european languages. Memorise some common patterns and structures but do not spend lots of time reading about abstract grammar rules.
Make hundreds of thousands of mistakes, get native speakers to make you aware of each and every single one, write them all down and drill them with a space repetition system like Anki. Find native speakers who arenât polite (this is quite hard in Chinese) and instruct them to correct you not only on pronunciation and grammar violations but on use of language which sounds slightly âoffâ and now how a native speaker would say it. If the language you are studying is Chinese and the native speaker is being honest with you, it is likely that at first every sentence you might wish to utter will be at least slightly wrong. Donât be put off, just drill the correct way of saying it and get it right next time.
Mass input is one key component among others NOT the end of the story. Input extremism is damaging. So is the idea that adults learn languages in the same way as Children acquire their L1. WE DONâT. The differences may have been exaggerated in the past but the swing in the other direction has been too violent and extreme. Will shows us that efficient adult language acquisition requires major intervention: space repetition, constant and relentless correction and feedback from native speakers. This is worth saying because there are people out there who think that because theyâve immersed in 100 novels that means one day they will be able to open their mouth and magically speak. This is deluded and tragic.
If your goal is to learn to speak and understanding the conversational language within a reasonable time frame then focus on learning the things YOU want to say and understanding things which are relevant to conversations YOU want to have.
Good points, Michilini!
However, each of the points you mention is worth discussing in depth, as none of them is as obvious as it seems.
Unfortunately, Iâm hard pressed for time right now because Iâve got to prepare a project involving various business processes so here are just a few brief remarks on your first point (letâs say as an appetizer for the near future):
The myth âearly output is dangerousâ is dead and buried.
Apart from certain input-oriented online groups (MIA, AJATT, Refold, etc.) early outputting is the norm both in schools and universities (not only in Germany, but also worldwide).
And this means: The lack of exposure to a lot of compelling and comprehensible content is still the main problem in these environments!
Krashen, Steve, Matt, and others would reply to you that the crucial thing in interacting with others is the input you get from them. So the deciding factor would still be input, not output
However (and thatâs my position based on advanced research reg. communication as social emergence):
The âinput/outputâ relationship and the corresponding âsender-receiverâ model per se are deeply flawed when applied to human communication and lead to nonsensical conclusions:
a) Input (listening/reading) = âpassiveâ and output (speaking, writing = active
b) And even better:
There is âactive and passive vocabularyâ and âactive and passive listeningâ.
A similar âlogicâ applies to the distinction âactive / passive vocabularyâ.
c) And as if that wasnât enough, there are more than enough people who seem to believe that communication is the transfer of âthoughtsâ from one mind to another mind or the âfusionâ of minds (âsocial mindâ, âintersubjectivityâ, âglobal consciousnessâ, âgroup consciousnessâ, yada, yada, yada).
The absurdities seem to know no end!
I donât intend to open that can of worms right now, but definitely in a few months when Iâve finished an updated version of German texts that I wrote a few years ago as parts of my PhD project.
Anyway, the only thing that is still valid right now is this: exposure to a lot of comprensible content both by immersion and social interactions is a must!
Everything else (i.e. communication models, Universal Grammar Ă la Chomsky, innatism vs constructivism, etc.) is open for discussion. And this means also that a lot of the old stuff (esp. the sender-receiver / IO model of communication applied to humans) has to be updated because weâre living in 2022, not in the 1950s any more
Related to 1) I tend to break things into âyou can do it now or do it later, but you have to do itâ. With learning you canât cram or force yourself to learn anything within a time, so you need to plant the seed far earlier than when you need it.
A learner of an L2, with the goal of having native like abilities in all areas, will have to make a lot of these decisions. They could start with everything from Day 1, but that is going to overwhelm almost anyone, at least for distance L2âs with alien writing systems.
So for most it will make sense to strategically stagger aspects of the L2. âParking Lotâ them or âplug holesâ with the intention of fixing it later.
Some might choose to ignore writing, some might choose to ignore gender, case, tones, or forego speaking altogether for a period of time. There is probably an âoptimalâ way to stack them, but IMO the order should come from whatâs important to the individual.
Where I think âspeaking earlyâ might become a âtrapâ is where we never try to address the things we put in the parking lot. If we become comfortable and stop trying to improve, those areas we ignored will stay unrefined. This is easily rectified by the individual, but, like everything else, it requires intention and deliberate effort.
Two more thoughts for food for the following days:
âdrill them with a space repetition system like Ankiâ
Extensive reading works like a ânatural SRS.â That is, we know from SLA research that extensive reading amazingly expands learnersâ vocabulary (not to mention the fact that reading is usually more interesting and motivating than using only artificial SRS)!
What we should compare is
not fluency first vs (ultra)reading while listening (after learning complex writing systems), for example,
but the whole package, i.e. listening + speaking + reading (while listening) + writing.
That is:
If someone adopts an MIA / input first approach, (s)he will still need to speak (and write) a few hundred hours to reach a high (or even native-like) level of speaking and writing. But his / her advantage is that speaking in particular will be a breeze.
If Will wants to have the full skillset as well, he still has the harder part ahead of him, i.e. learning thousands of Chinese characters, reading and writing a lot.
The comparison for the whole language package is therefore, e.g.:
fluency first + learning the writing system + (ultra)reading while listening
versus
learning the writing system first + (ultra)reading while listening + fluency second.
In other words:
Willâs approach is, in my opinion (so far), only the clear winner when two conditions are given:
Learners are happy to speak fluently (speaking + listening), but are also content to be illiterate.
There are complicated writing systems involved.
This would mean, for example, that in my next language experiment (Dutch) I expect Willâs approach to no longer have many advantages over â(ultra)reading while listening from minute 1 plus Fluency Secondâ.
In sum:
it depends on the specific use cases (goals, time constraints, etc.) and the distance of the L2s, especially the difficulty of the pronunciation and writing system(s), which mix of approaches is superior.
From my perspective, it probably boils down to these three use cases:
USE CASE 1:
Relatively âeasyâ languages (letâs say: Germanic / Romance languages for closely related Indo-European native speakers): when it comes to mastering all four language skills, it doesnât really matter if a âfluency first and then URL secondâ approach or âan URL first and fluency second approachâ is adopted.
USE CASE 2:
For distant languages with difficult pronunciation, an unfamiliar writing system and the whole skillset: fluency first Ă la Will and URL second is IMO the way to go.
USE CASE 3:
L2s between these two extremes,e.g.: Russian, Vietnamese (with its Latin script), etc. as L2s to be acquired by German, English, etc. native speakers - or vice versa: I donât know
Use case 1 is probably not interesting, but the decision for use cases 2 and 3 is:
Hard first (learning the writing system + URL), easy later (beyond conversational fluency)?
Easy first (conversational fluency). hard later (learning the writing system + URL)?
Hi Peter. First off, many thanks for listening to the podcast and for your detailed response.
Krashen, Steve, Matt, and others would reply to you that the crucial thing in interacting with others is the input you get from them. So the deciding factor would be still input, not output
First of all I donât think Steve or Krashen have ever said that early output is dangerous so this doesnât apply to them. Secondly, my point stands. If someone wants to say it is dangerous - despite the mountains of evidence to the contrary then the way to proceed is simple. Present the evidence.
Extensive reading works like a ânatural SRS.â That is, we know from SLA research that extensive reading amazingly expands learnersâ vocabulary (not to mention the fact that reading is usually more interesting and motivating than using only artificial SRS)!
Really. Have you tried learning characters to advanced literacy in a character based language? If so I have no idea how you managed it without some kind of space repetition intervention. In my experience, reading acts as a natural space repetition system for the most common words and characters. But in Chinese if you want to be literate you need 3000+ characters. Good luck getting from 2000-3000 without something like Anki. Some people use memory techniques but thatâs still an intervention. In either case immersion is clearly insufficient rendering your argument null and void.
In other words:
Willâs approach is, in my opinion (so far), only the clear winner when two given conditions are given:
Learners are content to speak fluently (speaking + listening), but are also content to be illiterate.
There are complicated writing systems involved.
Three words that instantly and completely demolish your argument here: Will. Can. Read. He hasnât practiced reading much, but nevertheless he can read. This is actually a very commonly reported experience among people who achieve excellent oral skills and drilled flashcards to learn characters while learning to speak. Their literacy skills advance at lightning pace because, like children, they already know the words. And, better than that, they already know the characters.
@michilini:
The âearly output is dangerousâ myth is dead and buried.
By itself maybe. But Iâm not sure you can tease out the combination of other factors. Unless youâre just saying that output is part of the mix. Iâll buy that with output the other factors are complemented. I donât by that output in and of itself is key. Benny Lewis showed it isnât when he attempted Mandarin. So other factors are necessary.
2. With Chinese, please for the love of god (if it isnât already too late) focus on tones and pronunciation from the start.
This is probably fair. And for distant languages in general (tonal or otherwise) that have very different pronunciation
than your native language.
Also with Chinese specifically, use a space repetition system that focuses on WHOLE SENTENCES and phrases rather than memorising thousands of words in isolation. Internalise the grammar in this way and DO NOT, under any circumstances, approach the challenge of Chinese grammar in a similar way to how you might approach european languages. Memorise some common patterns and structures but do not spend lots of time reading about abstract grammar rules.
Iâve been thinking about this one and to me this is the least solid conclusion of all. For grammar in particular, how do you know he didnât obtain the grammar from taking part in conversation (i.e. immersion) and passively from his massive input.
So while this might be true, there was no double blind control so itâs not statistically valid enough to be able to draw a solid conclusion. So Iâm not buying it.
I do buy the part about not spending lots of time reading about abstract grammar rules. Thatâs a given and most polyglots know this already.
Mass input is one key component among others NOT the end of the story.
While I agree with most of your points here I think you have missed something. Will spent a massive amount of time. Much more than most people would. It is not therefore legitimate to discount mass input in the way you have hand waved away. In fact, a large component of his success is immersion for a solid year every day pretty much every day for the first six months and several hours for the rest. He had a chinese girlfriend. He had/has chinese friends. I donât think you can say this doesnât count as mass input.
My personal theory about immersion is this: get dropped into prison. In a year youâll speak the language fluently.
If your goal is to learn to speak and understanding the conversational language within a reasonable time frame then focus on learning the things YOU want to say and understanding things which are relevant to conversations YOU want to have.
I donât agree with this at all. In fact unless you are immersed this will be counterproductive. You need to learn the most relevant frequency words or youâre done.
@michilini: But in Chinese if you want to be literate you need 3000+ characters
But not just the characters youâre interested in or like the look of right?
The 3000+ most frequent characters.
Michilini,
Do not write off other approaches. This is David Long who learned to speak Thai through listening and no speaking, writing, and reading in the meantime. He listened, listened and listened. He ended up speaking Thai almost as a native speaker with no prior speaking practice. However, it took him 30 months of listening 6-7 hours every day. Speaking emerged on its own. All he had done was listening and guessing like a child does. It is a long time and plenty of patience is required for this method to work.
If Will is honest about his timeline, what he has done is commendable.
Just watch this conversation with an open mind and see if you agree with Davids ALG approach which is based on input and observing silent period in terms of speaking output.
@Michilini, I largely agree with you, but I feel two points could use some clarification.
Regarding speaking: I would like to differentiate between:
1 making sounds:
For example using the shadowing technique. Starting to make sounds as early as possible, seems sensible to me: it trains the muscles, sharpens the hearing, reduces fear of speakingâŚ
2 talking to people:
Will is a great example of how this can work early on, but it needs the right conditions: kind, patient, willing native speakers, but also in the case someone like Benny Lewis the ability to discard fear of imperfection and being judged.
My opinion is that delaying 2 somewhat to improve oneâs vocabulary and listening comprehension can be beneficial. Understanding the other party gives you at least a chance at formulating a response. Thatâs why I plan to develop my listening comprehension to a âpracticalâ level (letâs say: understanding podcasts effortlessly) before bothering native speakers. This choice is of course a reflection of my own personality and circumstances and has no bearing on Willâs example. I would say Iâm closer to Steve Kaufmann on this matter than to antimoon.com / AJATT. The latter two are also opposed to 1 of course.
3 speaking to yourself / thinking in the target language:
I feel this is a good practice for the phase between 1 and 2. It doesnât require listening comprehension but still allows practicing the sentence structure, formulating thoughts in the TL and provides at least some feedback, albeit not from natives.
Re: SRS
I do not do Anki, nor have I ever. So whatever I know about characters, I have learned by writing them out by hand, watching subtitled videos or listening and reading on LingQ. I donât have an argument itâs just that I donât think Anki is strictly necessary. Lots of people on the internet seem to love Anki and are very successful, so go ahead everybody.
