Transcription
Yesterday I had a 30-minute conversation in Spanish with a native speaker, and 3 months ago I could not string together a sentence. No edits, no cuts, no preparation in the moment. Just 30 minutes of real conversation in a language I basically didn't speak 90 days earlier.
And here's the thing: the single technique that made the biggest difference is the one that 99% of language learners will never do. It's not immersion. It's not Duolingo. It's not just watch Netflix. And in fact, when I tell people this, most of them laugh. They tell me it's outdated. They've been told for years that this technique doesn't work, that it's old school, and that real fluency comes from input, exposure, and vibes.
And let me tell you, they're wrong. And I'm going to show you exactly why with the science, the history, and what happened when I actually tried it. So, let me give you a little bit of context. I just finished a three-month uh Spanish challenge, zero to conversational, ending with a real call with a native tutor on italki. And I'll be transparent, I'm a native French speaker, so I had a head start. But what took me 3 months might take you six. But the FSI, which is a US Foreign Service Institute, classifies Spanish as one of the easiest languages for English speakers. And so this is achievable for most people with a smart program.
On this channel, I've interviewed dozens of scientists and applied linguists about language learning that actually works. One of them is Professor Paul Nation, arguably the most cited researcher on second language uh vocabulary in the world. And buried inside the self-study book, *What You Need to Know to Learn a Foreign Language* is the technique I'm about to show you. He calls it activity 4.1, memorize sentences and dialogues. And it sounds boring. It sounds basic. It sounds like something your high school Spanish teacher would have made you do. And that's exactly why nobody does it and exactly why it's the most underrated language learning technique on the planet.
Now, let me describe a learner you probably know. Maybe it's you. They've been learning Spanish for 2 years. They watch Spanish language Netflix. They listen to Spanish podcasts on their commute. They have a 400 Duolingo streak. They've consumed thousands of hours of input. And this is the dirty secret of the immersion-only crowd: pure input by itself does not make adults fluent on a reasonable timeline. There's a reason for that, and it has nothing to do with talent. It has to do with how your brain actually stores language.
And here's what the research has been telling us for 40 years, and what almost nobody on YouTube wants to hear: Native speakers don't build sentences word by word. They retrieve them whole, pre-assembled, like Lego bricks already snapped together. And in 1983, two linguists named Andrew Pauly and Francis Cider wrote a paper that changed how applied linguists think about fluency. They argued that fluent native speakers have thousands, possibly hundreds of thousands of memorized sentence stems stored in their heads. So they don't construct "I was wondering if you could" from grammar rules. They pull it off the shelf.
Then the corpus linguists got to work. In a famous 2000 study, Emmen and Warren actually analyzed real speech and concluded that about 58% of spoken English is made of prefabricated chunks. Other studies pushed that number to 80% in casual conversation. So, play that part again: more than half of what comes out of a fluent speaker's mouth is not invented in the moment; it's retrieved from memory. So the question becomes obvious: if fluent speech is mostly memorized chunks, what exactly do you think you're doing when you refuse to memorize anything?
So now I want to talk a little bit about why all these memorization techniques are dismissed. Why is it that the modern language learning internet hates memorization? And there are two reasons. And the first is history. Back in the '40s and the '50s, there was a method called audiolingualism, and soldiers and students drilled memorized dialogues for hours with no real communication. And so by the 1970s, it had collapsed, partly because Chomsky destroyed the behaviorist theory underneath it, and partly because students who only drilled couldn't actually talk to anyone in the real world. And so memorization got tarred with the failure of audiolingualism, and the entire field swung the other way.
Second, Stephen Krashen. Now, in the '80s, Krashen proposed that comprehensible input, which is listening and reading to things you mostly understand, is basically the only thing that mattered. That hypothesis dominates YouTube polyglot culture today: "Just get input. Don't study. Acquisition, not learning." And here's the problem: Krashen's hypothesis has been challenged for decades by mainstream SLA researchers. Output matters, feedback matters, deliberate practice matters. And a 2025 paper in *Frontiers in Psychology* called the "input-only view a passive process that ignores how the brain actually rewires itself through sensory-motor engagement, social interaction, and continual feedback." So the translation here, in plain English, is that you don't learn to ride a bike by watching cycling videos. You don't learn a language by only listening and reading.
And Professor Paul Nation himself addresses this directly in his framework called the four strands. He argues a balanced learner should split time roughly equally between input, output, fluency practice, and deliberate study. And so memorizing sentences belongs in that fourth bucket. He doesn't say it should be the whole program, but he says that it should be about a quarter of it. And that quarter is exactly what most learners skip.
Now, let me talk about the science. I guess this is the fun part, the actual evidence that memorizing whole sentences and even dialogues actually works. In 2008, Conklin and Schmidt ran a famous experiment. They had native speakers and second language learners read either formulaic phrases or matched non-formulaic phrases that meant the same thing. And the result is that both groups read the formulaic phrases significantly faster. The brain processes pre-stored chunks more quickly than novel constructions, every single time.
This matters because of a thing cognitive scientists call working memory. So your working memory in your brain is small, tiny. It can hold maybe four or five things at once, depending on the person. When a beginner tries to construct a sentence in real time, pick the right vocabulary word, conjugate the verb, get the gender right, remember the word order, they exhaust their working memory in two seconds. And so that's why they freeze. Now imagine that you've memorized "I'd like to order" as a single unit. You don't compute it; you just retrieve it. Working memory free, ready for actual content.
And also in 2006, a team led by Frank Bowers ran a control study testing this directly. They taught one group of learners using a normal approach, another group using explicit chunk memorization, and then blind judges rated the speakers on perceived oral proficiency. The chunk group rated significantly higher, and they sounded more fluent. They sounded more native because they were retrieving instead of constructing.
So there is a a cognitive theory called skill acquisition theory, developed by Robert de Keyser. It says that any complex skill, like playing piano, driving, speaking a language, moves through three stages. First, you know the rule. Second, you can apply it slowly and consciously. And third, you can do it automatically without thinking. The only thing that gets you from stage two to stage three is repetition. Lots of it on the same units. That's exactly what memorizing sentences and dialogues is. And that's not even getting into, you know, spaced repetition research, retrieval practice, the testing effect. Every major finding from cognitive psychology in the last 30 years points to the same conclusion, which is that actively recalling chunks of language repeatedly over time is one of the most efficient ways to learn anything.
So we have the linguistics: native speech is mostly chunks. We also have the psycholinguistics: chunks are processed faster and free up working memory. We have intervention studies: explicit chunk memorization measurably improves perceived fluency. We have skill acquisition theory: automaticity requires repetition. And we have decades of memory research validating the technique. And against all of this, the modern internet says, "Don't memorize, just get the input."
If you want more proof, just look at what works in the real world. In World War II, the US Army ran something called the Army Specialized Training Program. They needed soldiers who could speak Japanese, German, Russian, and 30 other languages fast. Their program took complete beginners to functional speaking ability in about 3 months. And the core technique: relentless drill on everyday memorized phrases.
Something else as well: Mormon missionaries. We have plenty of them here. They train at the Provo Missionary Training Center. They take 18-year-olds with zero background and put them on a plane to Korea or Japan or Russia after nine weeks of training. The official manual instructs them to memorize simple statements that convey key gospel principles. Also, recently, the US military has visited the MTC to ask, "How are you doing this so fast?" There's a pattern here. Every program that has ever consistently produced fluent adult speakers on a tight timeline puts memorized sentences and dialogues at the center of the method. And on YouTube, the dominant message is, "Don't bother memorizing, just immerse." It's insane.
Okay, so how did I actually use this in my Spanish challenge? So when I started, I knew exactly what my final test was going to be: a 30-minute conversation with a native Colombian tutor. So I asked myself a simple question: What is going to come up in that conversation? Easy. Where I'm from, what I do, why I'm learning Spanish, my YouTube channel, the challenge itself, uh, my French background, my favorite shows, my family. So I sat down and wrote short speeches on every one of those topics, multiple speeches in Spanish, with help from AI to make sure that the texts were actually correct. Then I memorized them: whole sentences, whole patterns, whole transitions, whole speeches.
And a lot of people online are going to call this cringe. "You're just reciting. You're not really speaking." But here's what actually happened in the final conversations: I did not recite the speeches word for word. I never had to. What happened is that the patterns in those speeches, the connectors, the transitions, the ways of expressing opinions and time and cause and effect, it was already automatic in my head. So when the people I talked to asked me an an unexpected question, for example, about soccer with Gold Remy, I did not freeze. I had structures ready. The phrase "the truth is that" came out without thinking. The phrase "what I mean is," the phrase "for example" – none of these required computation; they were retrieved. And that's kind of the magic: you don't memorize speeches to recite them; you memorize them to internalize the building blocks so that when real conversation happens, you have spare working memory to to think about what you want to say, not how you want to say it.
And here's something that I did not expect: Memorizing those sentences also massively boosted my listening comprehension. Because once a phrase is automatic in your head, you instantly recognize it when a native speaker uses it at full speed. And the chunks become anchors that pull the rest of the sentence into focus.
All right, so how do you actually do this? Step one is identifying the situations you'll actually use the language in. So, be honest. Just, you know, you're learning Spanish for a trip to Mexico City, that's going to be restaurants, taxis, hotels, asking directions, maybe small talk. If you're learning Spanish for work, it's going to be introducing yourself, your job, your projects. If you're learning for a partner's family, it's family conversations, opinions, telling stories about yourself.
Step two, write short dialogues or speeches or sentences for every single situation. Multiple dialogues, multiple sentences for every situation, two to three minutes long. Get a native speaker or a competent AI tool to make sure that they sound natural. You know, the robotic textbook phrases don't really help you; you want phrases that real people actually say.
Step three is going to be audio. So, you're going to have to have a native speaker pronounce them or use very high-quality text-to-speech software. You need to know what they sound like before you memorize them, or you're going to lock in bad pronunciation for a long time.
So, what you need to do is you need to force yourself to retrieve the dialogue or the sentence in your head. This is the key step: retrieval, not just reading. The harder it feels, the better the learning. After you retrieve the sentence in your mind, you need to practice saying it out loud at full speed until it's really automatic. Do this every day, 15 minutes. You can do more, but really 15 minutes is good.
And step six, and this is what Paul Nation specifically warns about: Make sure that you also understand the parts. Don't just memorize as like a magic spell. Understand that "I would like" is "I would like," and you can swap in any verb. The whole point of chunks is that they teach you grammar through use, but you have to actually pay attention. And that's it. That's the whole technique.
So now, why is it that most people don't actually do this? And I think there are three reasons, in my opinion. Number one, it's unsexy. It doesn't trend on TikTok. There's no AI chatbot disrupting it. It's just you, a text, and a willingness to do something boring for 15 minutes a day, at least. Internet language learning culture has been telling people for 15 years that memorization doesn't work. They believe it. They never test it; they just keep watching Netflix in Spanish or whatever language they're try they're trying to learn and wondering why they they still cannot talk after three years. Number three, it requires honest self-assessment. You have to figure out what you'll actually need to say, and that takes some thought. It's much easier to passively let an app feed you content than to sit down and prepare.
But the 1% who do this technique – the diplomats, the missionaries, the polyglots, the people who walk out of intensive programs actually speaking – share one thing in common: They were willing to do something that was boring but it worked, instead of doing something fun that did not.
So, I'm not telling you to, you know, memorize sentences and skip and skip everything else. I had four other tools in my Spanish challenge: a course, vocabulary, flashcards, an AI tutor, and Netflix. They all matter. Paul Nation himself says that memorization is part of a well-balanced program. But really, that part, that memorization part, is really the difference between learners who plateau forever and the learners who actually become fluent. It's really the bottleneck that almost everybody ignores.
So, if you're going to take away one thing from this video, take this: Pick five situations you actually want to handle in your target language. Write a short speech or dialogue for every single situation. Memorize the dialogues. Do it for the next 30 days and watch what happens to your speaking. The science backs it. A century of effective language programs backs it. My own 3 months backs it. The only thing standing between you and the result is whether you're willing to do something the internet told you not to. Most people are not going to do it. And that's why most people will never speak a second language fluently, no matter how many years they spend learning one. But you can be different and start today. I'll see you in the next one.