⏳ Method — the question of total hours
A lawyer asked us this over breakfast. He wanted the international work at his firm, so he was taking four half-hour conversation lessons a week over video and a two-hour class with a teacher at the weekend, and after months of it nothing had moved. He is not a man who cannot study — he passed a bar exam. The answer turned out to be arithmetic rather than neuroscience.
The Foreign Service Institute teaches languages to American diplomats, and it publishes how long each one takes. Category I — Spanish, French and their neighbours — runs 24 to 30 weeks, or 600 to 750 class hours. Category II, where German sits, is about 36 weeks (900 hours). Category III, Russian and Turkish, about 44 weeks (1,100 hours). And Korean, Japanese, Chinese and Arabic sit in Category IV: 88 weeks — 2,200 class hours, filed as exceptionally difficult for a speaker of English.
These are full-time students, selected for aptitude, in very small classes with professional teachers, doing homework on top of the class hours. So the figure is not “put in this many and you are done”. Read it instead as a measurement of how far apart two languages sit. And the distance does not shrink when you run it the other way: there is no official timetable for a Korean speaker learning English, but the gap between the two languages is the same gap seen from the other side.
Worked out exactly as he described it.
If it only accumulated, ten years would at least get there in the end. The problem is that it drains. What Hermann Ebbinghaus showed in 1885, memorising nonsense syllables on himself and plotting what was left, is simple enough: without review, what you learned leaves fastest in the first few days. (His material was nonsense syllables, so the percentages do not transfer to sentences in a real language. The only thing to carry across is the shape of the curve.) Four hours a week is barely enough to patch that hole, let alone fill the bucket above it.
Add hours, or change what happens inside the hours you already have. For anyone with a job the first is mostly closed — which is why we went at the second. Raise the playback speed so that more sound passes through the same ten minutes, and let what passes become familiar rather than memorised. We do not attach a multiplier to that; we have never measured one, and any number we put here would be invented. The one thing we are sure of is the direction: if the hours cannot grow, density is the only handle left. Why the sound goes past you at all is the other half of this story ([[why-english-doesnt-click]]); the order that makes density possible is in [[listening-first-order]], and the drill itself in [[what-is-shadowing]].
Start with a language you already understand and play it faster than normal. Meaning costs you nothing, so every bit of attention goes into the speed at which your brain handles speech — and that speed carries across when you switch to the language you are actually learning. Changing the order is itself a way of changing the density. (If the language you are learning is Korean, our free course is built on this order: brain-hz.com/korean — 60 lessons, shadowing and speak-first modes, uncompressed audio, free and no account.)
None of this says that conversation lessons are wasted or that a teacher is unnecessary. The lawyer’s four hours were good hours. They were simply being spent on output at a stage where the bottleneck was intake — and no amount of speaking practice, at four hours a week, closes a gap that is measured in thousands.
Source
The category ratings and class hours are published by the Foreign Service Institute of the United States Department of State (state.gov) — Category I 24 to 30 weeks (600 to 750 class hours), Category II about 36 weeks (900), Category III about 44 weeks (1,100), Category IV 88 weeks (2,200: Korean, Japanese, Chinese, Arabic). The forgetting curve is Hermann Ebbinghaus, Über das Gedächtnis (1885). This is the arithmetic half of [[why-english-doesnt-click]]; on the target worth aiming at instead, see [[standard-english-not-native]]. — Not carried over from the Korean edition: the original blog post’s “3,000 hours needed for English”, its “900 to 1,000 sentences of input in ten minutes”, and its claims of being ten to a hundred (or five hundred) times ordinary study, none of which we could trace to a primary source. The baseline here is the published figure instead.
Read the original →※ This article summarizes and restructures the key points of the source above.
🧭 How the method works · 11 more on this