3 Reasons You Should Learn Languages the Natural Way - Steve Kaufmann

:brain: The brain acquires languages through lots exposure of input, NOT the conscious memorization of grammar rules and vocabulary. This is good news! In this video, I share 3 major benefits of learning languages in a natural, non-rules-based way.


1 Like

This is true but a hyperbolic expression of the truth.

I bought into the idea too much myself and am now going back to take grammar lessons to fill in holes.


I don’t believe Steve is against learning grammar. In other videos, he encourages looking things up if they come up and you are curious. I think the difference is studying grammar and trying to memorize it vs. reading and looking up things in the context of what you have read for better understanding.


Indeed. I believe the following oversimplifies it if not even is misleading considering the YELLING.


I don’t know. I tend to agree with Steve’s thoughts on this. Even for my own native language…the vast majority of grammar was not learned by memorizing grammar rules. In fact, I don’t recall grammar being taught really beyond the elementary grades. I certainly wasn’t doing grammar exercises beyond elementary levels. Had I learned ALL the grammar by then. No way. So how did I learn it? From reading mostly…listening. From writing (and getting corrections)–of which I didn’t get a grammar rules written out in the red ink by the teacher, but the correct word or usage of the word or phrase.

When someone says something that is grammatically wrong or asks for help with their English (or other native language), do we recall a grammar rule? Most of us don’t…it just sounds wrong and we say to them how it should be. Why does it sound wrong? Because we’ve heard and read the correct way hundreds or thousands of times.

Even if we look up a gramar rule. We don’t really memorize the rules, we look at repeated patterns of that grammatical context over and over again. Or we write out the pattern over and over again in the exercises. It’s the pattern that’s sinking in. Not the rule.

That’s not to say there isn’t benefit to looking up grammar topics, but it’s not the rule you memorize, but rather the patterns in the examples they provide. Thinking of a rule is going to slow you down. I think there is benefit too, to looking up a grammar item if you come across it in your reading and want to know a litle more about why it is that way. Then you have a little more context to the situation that may help the grammar sink in.

True…maybe the description of the video is a little click-baity, but I don’t think it’s too far off.

Ya. Just had a little bile in the mouth from the click-baity-ness.

And maybe you’ve emphasized my own point on your reference to “hundreds or thousands of times.”

Even with the hundreds of hours of reading and listening to French as a second language, and even with now having a 33,000 word vocabulary in it at this point, there are still irregular conjugations that I’ve not yet enough times to master them by immersion. Further, with French and its non-phonetic nature, it’s important to listen much and read much as one needs to internalize the patterns of both pronunciation and orthography with a decent number of exceptions to the rules.

Learning the grammar rules–or better as you identify, “patterns”–directly can offer a higher efficiency than consuming yet more and more volumes of native content that may only use certain forms quite infrequently.

Now, perhaps, I do need to say that testing and certification are indeed real-world concerns of second language learning. Perhaps I’m being a bit jaded, but part of the job of a “test” for any credential is to help create the demand for the purchase of training services. I work in a tech field and they put the most obscure questions on tests that I swear are there just to test whether or not you went through the paid training program! I do not believe the second-language-learning business is immune from these commercial forces.

If one is trying to pass a certification test as a very significant waypoint in the overall success strategy second-language communication goals, a diet of comprehensible input alone may not be the fastest or cheapest or least stressful way to get there.

Summarizing, while I am a HUGE advocate of “the natural way,” saying that the brain does NOT acquire language through conscious memorization of grammar rules and vocabulary is symptomatic of a zealotry of a method that has to bear the burden of at least a small irony with its iconoclastic framing of “the traditional method.”

Now, getting more practical about it as I kick my soapbox out from under my own two feet, it would be extremely powerful if LingQ helped me better learn grammar.

Here’s a simple, content-centric way your current UI could be extended to support grammar learning.

Take this sentence of French content from yesterday:

“Continuant son cours de droit en accéléré, il a rappelé que lorsque la Charte est invoquée, le procureur général doit être mis en cause, ce qui n’a pas été fait.”

(Which LingQ translates to English as, “Continuing his crash course in law, he reminded us that when the Charter is invoked, the Attorney General must be implicated, which was not done.”)

I can ask ChatGPT such as, “Enumerate the verbs in the following phrase and for each, identify their conjugation: ‘Continuant son cours de droit en accéléré, il a rappelé que lorsque la Charte est invoquée, le procureur général doit être mis en cause, ce qui n’a pas été fait.’”

The LingQ system knows which words in the sentence the user has marked as unknown. Often, you already have tagged the parts of speech and conjugation. With generative AI, there are ways you can do this much more efficiently. With a bit of your system’s integration with ChatGPT or similar behind the scenes, you can develop much more insight into your users progress with grammar.

In this specific sample sentence, there’s a lot of passive voice. Let’s say that’s something that I’m struggling with. Also, note the irregular verb “faire” conjugated as “fait.” Right now, when I click on the word “fait,” I see LingQ struggles with knowing whether or not it should be mapped to the third-person present tense conjugation of the verb “faire” (“to do”) or the homograph noun “fait” (a “fact”).

Until recently I believe you had two unsurmountable blockers for an input-centric language learning platform like LingQ to seriously graft in grammar learning into the experience in a new, contextual way.

I believe the one blocker (tech limitations) is completely gone with the recent advances in generative AI.

I can, for instance, give generative AI a text (such as that of a LingQ lesson) and ask it to do such as: “Give me a 25-question multiple choice quiz for the conjugation of those tenses of those verbs using the vocabularies and key concepts of the text.”

Second-language learning is a generative AI use case par excellence. LingQ, with its content-centric approach is by far better positioned than any other language-learning platform to take advantage of the next wave of opportunity.

Rather than “conscious memorization of grammar rules and vocabulary,” consider that with smart use of generative AI grafted cleanly into the user experience, you can offer “contextual insight into grammar rules and vocabulary.”

Here would be a really simple one to code. As LingQ is now, when the user clicks on a word in a lesson, you provide (crowd-sourced I think?) dictionary definition of the word. Imagine, if you also asked ChatGPT a question behind the scenes such as what I provide example of here using the reference phrase above.

Here’s my ChatGPT prompt:

Describe in English as well as in its own language the usage of the word “fait” in the text, “Continuant son cours de droit en accéléré, il a rappelé que lorsque la Charte est invoquée, le procureur général doit être mis en cause, ce qui n’a pas été fait.”

Here’s an excerpt how ChatGPT responds:

In the text, the word “fait” is used as part of the phrase “n’a pas été fait.” In this context, “fait” is the past participle form of the verb “faire,” which means “to do” or “to make” in English. The phrase “n’a pas été fait” translates to “was not done” or “has not been done” in English. It is used to indicate that a specific action or task has not been carried out or completed.

Imagine if that, too, were available when simply clicking on a word in LingQ.

Honestly, very little code would be required to pull off just this one incredibly powerful feature to help in learning more than vocabulary, in the context of comprehensible input.

I bring this up this thread simply because your zealotry of approach may be a second blocker yet to be overcome.

LingQ already doesn’t actually use “the natural way.” It uses an “augmented natural way” where the augmentation is chiefly in vocabulary learning (powered by basic string matching). With generative AI, you can pragmatically augment the learning experience in new, innovative ways.

Beyond vocabulary, each of grammar, orthography, register, style, and more are within reach now for augmentation of “the natural way.”

The best is yet ahead.

As a bonus, here’s another prompt you could use behind the scenes: "What are the idioms used in the following text; for each, provide a definition, describe how they’re used in the text and more such as their register, social appropriateness (or not), giving two other examples of using the idiom, using the content area of the text to the extent possible. Here is the text to analyze: "

Asking ChatGPT to analyze one LingQ lesson I read produced such as the following which further illustrates the potential through addition of new features that never require complete departure from the content’s immersive experience.

“Écarter les demandes vaines”

  • Definition: This idiom means to reject or dismiss futile or baseless requests or demands.
  • Usage in the Text: In the text, it’s used to describe the action taken by the tribunal in quickly rejecting the antivaxxer’s claim. It indicates that the tribunal efficiently filters out frivolous claims.
  • Register and Social Appropriateness: This idiom is formal and appropriate for legal contexts. It conveys the idea of efficiently handling baseless claims in a formal setting.
  • Other Examples:
    1. Le juge a décidé d’écarter la demande vaine de l’accusé, car elle ne reposait sur aucune preuve solide. (The judge decided to dismiss the defendant’s baseless request because it had no solid evidence.)
    2. L’avocat a conseillé à son client d’éviter de présenter des demandes vaines devant le tribunal. (The lawyer advised his client to avoid making frivolous requests in front of the court.)
1 Like

Here’s the thing though: people with a knack for linguistic abstraction learn the rule and get the patterns for free. We’re not merely “memorizing” rules, we are “grasping” them, and we pretty comfortably recognize instances of that rule when we see them, without having to think about consciously applying the rule.

When you are comfortable with abstract rules, you learn them and then you kind of “feel” or “see” them in action - you don’t consciously think about them as if they were tools to decode the language, or recipes for generating sentences that you need to consciously access.

So for us, not familiarizing ourselves with the grammar of the language is asking us to reinvent the wheel every time we look at a language. Why would I painstaking build up an intuitive, non-self-conscious familiarity with the patterns of a language through trial and error, when I could spend a short amount of time consciously digesting the abstract structure to which the whole of the language conforms, and then get on with experiencing those patterns in action?

Not everyone likes doing it, but the top-down approach leverages everything you already know about language and radically shortens the time to looking at real material.

As an experiment I tried doing the Russian the pure-Kaufmann way with Lingq and I absolutely hated it. I felt like I was deliberately playing the game of language acquisition with a huge self-imposed handicap.


We can’t use the way we learnt our mother tongue as the model. I spent two years living in Montreal, often talking in French, and my grammar never progressed beyond the B1 level that I had when I arrived. And I didn’t learn much vocabulary.

I’ve met countless people who learnt English (or another language) when young. Those who arrived in the country before they were ~12 had no accent, and native level grammar. Those who arrived later had excellent skills, but a noticeable accent. I can think of an Iranian and a Korean in the latter group. I know a Lithuanian who speaks Russian (his father’s language), Lithuanian and Polish (his mother’s language). Even after ten years here his English is hard to understand due to poor accent and grammar. So clearly something changes as we grow older.

1 Like


I used Duolingo for French. I was shocked to discover that words I’d spend weeks ‘learning’ disappeared from my memory after a couple of months. The reason? Because I repeated the word countless times over a week or two, with zero subsequent repetition. That’s not spaced repetition. It was also full of errors in the English grammar, as well as Americanisms which I did not understand. (I haven’t had that problem with LingQ.) I was also unimpressed with the weird characters assumed by the actors, such as a suicidal woman, and a very snobby man. I have listened to hundreds of hours of French podcasts, and spoken to many French people, and never heard anything like that. I really believe those weird voices are a bad idea. What I needed was real speech, people talking naturally, with pauses and mistakes. That is the only way you learn pronunciation. I also tried Duolingo for German from scratch. Awful, really awful. I hated it.

I found Babbel a good introduction to German, and in a few months it brought me to a level where I can work through the German short stories here on LingQ. Maybe they are not perfect, but I’m making more progress with them, than with other methods. Verbs and grammar are starting to sink in. I use them alongside Anki. I put the complete sentences into Anki, and then review than each day, which is helping me immensely. In a sense I’m using Anki as a comprehensible input tool, allowing me to review the text in the short stories in a random manner over time.

My impression of LingQ for German from the point of view of a relative beginner is positive. I’m apprehensive about my next steps after the short stories, but such is life.

1 Like


My comments only applied to Duolingo French and, to a lesser extent, German. For German I found progression horrendously slow. I worked out that if I did 1 hour a day, it would take four years to finish, and I still wouldn’t be able to follow podcasts. I got mindlessly bored.

I agree LingQ German is not suitable for beginners. I was surprised by the number of bugs, but in their defence they are quick to fix them. Most Duolingo bugs were not fixed, and those that were took 6 months. I used it for eight months. It will take me at least six more months to assess LingQ.

I think LingQ could be suitable for beginners if they created beginner level short stories.

We seem to have discovered the same way to use Anki. :slightly_smiling_face:

The ‘holes’ exist because you didn’t get enough input. However much we think is enough is usually nowhere close to enough.

1 Like

I think that spaced repetition is an enhancement of the old technique of learning by flashcards. Anki is a good example.

However, use premade high-quality Anki cards to save time. There are many available online.

Some people share their decks on:

Or you can look into some specialized ones: