Daniel wrote in with a bit of a regret this week — he picked up a Portuguese grammar book way too young, spent a summer trying to decode verb tables instead of just... talking to people. And when he finally did open his mouth and try speaking to locals, that was the part that clicked. The part that was actually fun.
The book killed the thing before the thing even started.
Right. And he's been chewing on that contrast ever since, especially now that he's thinking about what he wants to pass on to Ezra. He's got five questions folded into this, really. Why do kids pick up languages through play while adults get buried in conjugation drills? Why did Israel succeed in reviving Hebrew as a living language while Ireland failed to do the same with Gaelic? What's actually going on with audio-first methods like Pimsleur, and why should that approach be locked behind a paywall? And then the one that I think is the most interesting — can you learn to speak a language like Greek, with a foreign alphabet, well enough to communicate without ever learning to read it?
That last one is the question behind the question. He wants to know if the alphabet is a gatekeeper or just... furniture.
So let's unpack what actually works when you're trying to learn a language, and what the failed experiments teach us about how brains actually do this.
The thing to name right at the top — this is not a how-to episode. We're not going to tell anyone which app to download. What we're really looking at is what the competing methodologies reveal about how the brain processes language. Because the gap between Daniel's instinct — put down the book, just speak — and the institutional reflex to start with grammar and writing, that gap is where all the interesting stuff lives.
And it's not just a preference thing. There's a mechanism underneath it.
There is. And the arc I want to trace goes like this. We'll start with why children acquire language the way they do — through play, through interaction — and what that tells us about the adult brain's approach. Then we'll look at the two great national experiments in language revival, Hebrew and Irish, because those are the cleanest case study you could ask for. One built a community of speakers. One built a school subject. After that, we get into the audio methods and the alphabet question.
So where do we start — the child brain versus the adult brain?
The critical period hypothesis. This is the idea that there's a window for language acquisition that starts closing somewhere around age twelve. After that, learning a new language becomes markedly harder. And the popular version of this is "children's brains are sponges" — they just absorb everything effortlessly.
Which is wrong in a specific way.
It's wrong in the most important way. The mechanism isn't a magical biological window where the brain is just more absorbent. What children actually have is massive contextual input. A typical five-year-old is getting something on the order of ten to twelve thousand hours of spoken language exposure per year. Every interaction is low-stakes. Nobody's grading them. Nobody's correcting their subjunctive. They try something, they get a response, they adjust. They try again. That's the engine.
So it's not that the sponge is better. It's that the sponge is submerged.
And the adult learner does the reverse. An adult walks into a language class and the first thing they get is a conjugation table. Present tense, past tense, future tense. Here are the rules. Now apply them. This creates a cognitive bottleneck, because you've shifted from pattern recognition — which is what the brain is actually optimized for — to explicit rule application, which is slow and fragile.
You're running the language through the prefrontal cortex instead of through the basal ganglia.
That's... yes, that's the clinical version of what I was about to say. The basal ganglia are involved in procedural memory — habits, patterns, things you do without thinking. When you speak your native language, you're not assembling sentences from rules. You're reaching for patterns you've used ten thousand times. The adult learner who starts with grammar is trying to do assembly, and it's slow, and it breaks under pressure.
So the child isn't learning conjugation. The child is hearing "I go, he goes" ten thousand times and the pattern just... settles.
And here's the thing — children do get corrected, but not in the way adults do. A child says "I goed to the store" and the parent says "you went to the store?" — not "that's the irregular past tense, let's review the paradigm." It's a reformulation embedded in the conversation. The child hears the correct form in context and adjusts. That's pattern inference, not rule learning.
So that's the mechanism. Which brings us to the two great national experiments. One that followed the child model, one that followed the adult model.
And the results are about as stark as you could ask for. Let's start with Irish. Daniel grew up in Ireland, he lived this. The Irish language has been compulsory in Irish schools since the foundation of the state. Every student studies it for something like thirteen years. The curriculum is heavily grammar-focused, heavily exam-driven. You learn to parse a sentence, you learn the rules of conjugation, you pass the test.
And then you never speak it again.
The 2022 census data tells the story. About forty percent of the population reported being able to speak Irish. But when you ask how many use it daily outside the education system, it drops to one point seven percent. One point seven. Thirteen years of instruction, and fewer than two in a hundred actually use the language in their daily lives.
That's not a revival. That's a museum exhibit.
It's a language that's been preserved in formaldehyde. And the reason, I think, is exactly what we just described. The Irish state treated the language as a school subject — a body of knowledge to be examined, not a thing to be lived. You conjugate the verb, you get the points, you move on. There was never a community of speakers waiting on the other side of the exam.
Now contrast with Hebrew.
Eliezer Ben-Yehuda arrives in Palestine in 1881. Hebrew at that point had been a liturgical language for nearly two thousand years. Nobody spoke it at home. Nobody bought groceries in Hebrew. It was a language of prayer and scripture, period. Ben-Yehuda's strategy was not to write a textbook and hand it to the schools. His strategy was to create a community of speakers. He insisted that his own family speak only Hebrew at home. His son Itamar Ben-Avi, born in 1882, was the first native Hebrew speaker in nearly two millennia.
Which is an extraordinary thing to even say. Two thousand years without a native speaker.
Ben-Yehuda reportedly refused to let his son be exposed to other languages in the home. His wife spoke Russian, French, and Italian, and he insisted — Hebrew only. There are accounts of him getting furious when he caught her singing lullabies in Russian. This was a linguistic hostage situation, but it worked.
I said that once. The hostage situation thing.
You did, and you were right. The point is, Ben-Yehuda understood something that the Irish language revival never grasped. A language lives in the mouths of children. If you don't create a generation of native speakers — children who acquire it through play, through interaction, through the same massive contextual input we just described — you don't have a living language. You have a curriculum.
And the Hebrew revival scaled that. It wasn't just one family.
Right. Ben-Yehuda's project expanded through the school system, but the schools taught in Hebrew — they didn't teach Hebrew as a subject. By the time the state was founded in 1948, there was already a generation of native Hebrew speakers. The language had been re-vernacularized. It was the language of the playground, the market, the argument in the street. That's the difference.
So the Irish model said: learn the rules, and someday you'll be able to speak. The Hebrew model said: start speaking, and the rules will follow.
And one of those worked.
The other produced a country where forty percent of people say they can speak a language and fewer than two percent actually do. That's a ninety-six percent attrition rate from the classroom to the street.
I hadn't done that math.
I just did it in my head. It might be slightly off.
It's in the ballpark. And it makes the point. So that's the community question. But Daniel's other questions are about what you do when you don't have a community — when you're an adult trying to learn Greek on your own, and you're staring at a grammar book or an audio course.
Let's talk about Pimsleur.
The Pimsleur method is audio-first, and it's built on a few specific principles. Spaced repetition — you hear a phrase, you're prompted to recall it at gradually increasing intervals. High-frequency vocabulary — the first few hundred words that do most of the heavy lifting in conversation. And forced production — you're not just listening, you're required to speak out loud in response to prompts. No writing, no grammar explanation, no conjugation tables.
And the research on this?
It works well for building oral fluency in the first five hundred to a thousand words. People who do a full Pimsleur level can typically handle basic transactional conversations — ordering food, asking directions, simple introductions. The problem is it plateaus. The method doesn't provide the massive input that children get. Thirty half-hour lessons is fifteen hours of audio. A child gets that much exposure in about a day and a half. You can't compress ten thousand hours of contextual input into a course.
So it's a starter engine, not a full vehicle.
And the pricing reflects that, but not in a good way. Pimsleur courses run about a hundred and fifty to three hundred dollars per level. Each level is thirty half-hour lessons. To get through three levels — which might get you to a functional tourist level — you're looking at somewhere between four hundred and fifty and nine hundred dollars.
For fifteen hours of audio content.
For audio content built on a technique — graduated interval recall — that is not patentable. The underlying method is public domain. The production values are real — you need native speakers, careful pacing, good audio engineering — but the technique itself is not proprietary. And yet there's no equivalent free or open-source alternative at scale.
Why not?
Because production cost is non-trivial, and nobody's figured out a funding model for it. You can't just generate this with a script and a text-to-speech engine — the pacing matters, the natural pauses matter, the native speaker intonation matters. Someone has to produce it, and that costs money. The open-source community hasn't cracked the "pay for studio time" problem.
The monopoly isn't on the technique. It's on the willingness to actually produce the thing.
On the marketing. Pimsleur has been around since the nineteen sixties. They have brand recognition. They have placement in airport bookstores. They're the default answer when someone says "I want to learn a language in the car."
Which brings us to Daniel's alphabet question. Can you learn Greek well enough to speak it without learning the Greek alphabet?
The short answer is yes. Oral fluency and literacy are neurologically separable. We know this because illiterate adults in oral cultures demonstrate full grammatical competence in their native language. They can speak with complete syntactic complexity, they understand register, they can tell jokes, they can argue. They just can't read or write. Literacy is not a prerequisite for fluency.
The alphabet is not the gatekeeper.
It's not. But there's a catch. Without the alphabet, you can't use a dictionary. You can't look up a word you heard. You can't read a sign. You can't text someone. You're confined entirely to oral transmission, which limits your vocabulary growth and makes error correction harder. If you mishear a word and you can't look it up, that error just... stays.
The alphabet is a tool, not a prerequisite. It's the search engine into the language.
That's the best way to put it. Learning the alphabet gives you lookup access. Skipping it gives you faster initial speaking fluency but a hard ceiling — you'll eventually hit a wall where you can't grow your vocabulary fast enough through listening alone.
What's the optimal path, then?
I think the evidence points toward audio-first for the first two hundred to three hundred phrases. Get your mouth around the sounds. Build the procedural fluency — the basal ganglia stuff. Then introduce the alphabet as a tool, not as a prerequisite. The mistake is starting with the alphabet, because then you're back to the adult model — you're decoding symbols before you've ever produced a sound.
Daniel's instinct to put down the book is neurologically sound for the first phase. The book becomes useful later. The error is starting there.
The error is also treating the book as the thing itself rather than a reference tool. A grammar book is a map. You don't learn a city by memorizing the map. You walk around, you get lost, you ask for directions, and you consult the map when you need it.
The map is not the territory. We've arrived at a cliché, but it's the right one.
It's the right one because it's true. And the Greek case specifically — Daniel mentioned Greek, and I think it's worth saying: the Greek alphabet is not actually the barrier. It's twenty-four letters. You can learn the Greek alphabet in a weekend. The barrier for most adult learners of Greek is the same as for any language — the lack of an oral community. If you're learning Greek in Dublin or Jerusalem, you don't have a Greek-speaking community around you. That's the real scarce resource, not the alphabet.
That loops us back to the Hebrew-Irish contrast. The scarce resource is always community. Ben-Yehuda built one. The Irish system assumed one would emerge from the classroom, and it didn't.
Which is where I think Daniel's framing has a blind spot. He's asking about methods — Pimsleur versus grammar books, audio versus text — and those matter, but they're secondary. The primary variable is whether you have a reason to speak the language with real people.
Hilbert: It's not the method.
Go on.
Hilbert: I spent three years in the nineteen nineties selling Berlitz courses door-to-door in Cork. I sold language courses to people who never opened the books. The ones who actually learned — the ones who came back and told me it worked — they were the ones who had a reason. A girlfriend in France. A job transfer to Germany. A holiday planned and they wanted to not be the tourist who points at things. The materials were almost incidental.
You were selling the promise, not the method.
Hilbert: We were selling a set of cassette tapes and a workbook for about three hundred pounds. The tapes had pauses for you to repeat — it was Pimsleur before Pimsleur was a brand. Same idea. Graduated recall, native speakers, the whole thing. The tapes were fine. The problem was that most people did the first three lessons and stopped. The ones who kept going had a person on the other end waiting to hear them try.
Daniel's right about the book, but he's wrong about Pimsleur being the answer.
Hilbert: The real monopoly isn't on audio courses. It's on immersion. You can't buy a community of speakers. That's what Ben-Yehuda built, and that's what Ireland never managed. Every language app is selling you a simulation of conversation. It's not the real thing. It's a treadmill with pictures on the wall.
Hm.
Hilbert: I had a customer — a man in his fifties, learning Italian. Did the whole course, all three levels, never missed a day. Went to Rome, walked into a café, and froze. Couldn't order a coffee. The tapes had taught him to respond to prompts, not to talk to a human being who might say something unexpected. The simulation broke on contact with reality.
That's the plateau we were describing. The method gets you to a certain point and then you need something it can't provide.
Hilbert: The something is another person who doesn't care about your accent.
What's the thing you'd tell someone who wants to learn Greek but doesn't want to learn the alphabet?
Hilbert: I'd tell them the alphabet is the easy part. Twenty-four letters. You can learn it on the flight over. The hard part is finding a Greek person who'll talk to you slowly for six months. If you've got that, you don't need my cassette tapes. If you don't, the alphabet won't save you.
That's... actually the cleanest summary of this entire episode.
Hilbert: I sold the tapes. I'm not proud of it.
You seem fine with it.
Hilbert: I kept a set of the Italian ones. They're in a box somewhere. Never learned Italian.
The misconception I want to name before we close — there are really three that run through this whole discussion. The first is that children learn languages because their brains are sponges. It's not a magical biological window. It's massive contextual input in low-stakes environments. The second is that Irish revival failed because the language is too hard. It failed because it was taught as a school subject rather than built as a living community. And the third is that you must learn the alphabet before you can speak. Oral fluency and literacy are neurologically separate. You can speak without reading. The alphabet is a tool, not a prerequisite.
The thing all three misconceptions share is that they mistake the method for the mechanism. The mechanism is always interaction. The method is just how you get there.
The open question we're left with — if community is the real scarce resource, can technology actually create one? Not simulate one — create one. VR language spaces, tandem apps that pair learners with native speakers, the kind of thing where you're not talking to a chatbot but to a real person who happens to be in another country. Do those scale, or are they just higher-fidelity simulations?
I think the honest answer is we don't know yet. The technology is improving faster than the research can track. But I suspect the thing that makes community work — the low-stakes, ungraded, just-messing-around quality of children's play — is really hard to replicate in a structured app, no matter how good the latency is.
For Daniel and Ezra, I think the best gift isn't a method at all. It's an environment. A reason to need the language. If Ezra grows up in a multilingual household, in a multilingual city, with parents who model that languages are for talking to real people — that's worth more than any course.
This has been My Weird Prompts. Thanks to our producer Hilbert Flumingtop, who apparently has a box of Italian cassette tapes and no intention of ever using them.
If you've got a weird prompt you want us to tackle, send it to show at my weird prompts dot com. We'll be back soon.
