Ask how long it takes to learn a language and you’ll get a number delivered with more confidence than it deserves. Six hundred hours. Two years. Ten thousand hours, from someone who half-remembers a Malcolm Gladwell book about violinists.
Nearly all of it traces back to two sources, both routinely misread. Here’s what they actually measured.
Sources are linked inline so you can check every claim yourself.
Source one: the FSI category system
The US Foreign Service Institute trains diplomats to work in foreign languages, and it has decades of records on how long that takes. Its language difficulty categories are where essentially every "hours to fluency" figure on the internet comes from, usually uncredited.
Roughly, for a native English speaker:
- Category I: about 600–750 hours. Spanish, French, Italian, Portuguese, Dutch, Norwegian, Swedish, Danish, Romanian.
- Category II: about 900 hours. German, Indonesian, Swahili, Haitian Creole.
- Category III: about 1,100 hours. Russian, Polish, Czech, Greek, Turkish, Hebrew, Hindi, Persian, Thai, Vietnamese, Tagalog.
- Category IV: about 2,200 hours. Arabic, Mandarin, Cantonese, Japanese, Korean.
The numbers are real. What gets stripped out when they’re quoted is the problem.
Four things people leave out about the FSI numbers
Each of these makes the number less applicable to you. They compound.
- The target is professional proficiency, not chatting. FSI trains to roughly "Speaking-3 / Reading-3" on the Interagency Language Roundtable scale, somewhere around upper B2 to low C1. That’s the level where you can negotiate policy in the language. Ordering dinner and holding a decent conversation arrive much, much earlier.
- These are full-time, intensive, classroom hours. Six students, a native-speaker instructor, a structured curriculum, all day every day. One of those hours is not equivalent to one of yours on the train.
- The students are unusually strong. Screened adults, many of whom already speak two or three languages, doing this as their salaried job with a posting depending on the result.
- The hours count class time only. Homework and evening study sit on top of the figure, uncounted.
The FSI figures describe how fast a screened professional reaches working proficiency under ideal full-time conditions. Read as a promise about your evenings, they are actively misleading.
Source two: CEFR guided learning hours
The other common numbers come from estimates attached to the Common European Framework of Reference levels, the A1-to-C2 scale. Various exam boards publish approximate guided-learning-hours ranges: very roughly 80–100 hours to reach A1, a few hundred to B1, and something like 1,000–1,200 cumulative hours to reach B2 for a European language.
Two caveats. These are institutional estimates rather than experimental findings, and publishers disagree with each other. And "guided learning hours" again means taught time only.
What CEFR gives you that FSI doesn’t is a ladder. Not one distant number, but checkpoints, and the early ones arrive fast enough to keep you going.
Why "fluent" is the wrong unit
The real trouble is the word "fluent." It has no agreed definition, which means any hour figure attached to it answers a question nobody actually asked.
Ability isn’t one dial. I know people who read Borges comfortably and freeze ordering coffee. Ask what you want to do instead:
- Order food, ask directions, handle a trip: weeks to a few months of focused effort.
- Hold a real conversation with a patient speaker: several months to a year.
- Follow native media without subtitles: considerably longer, and listening is usually last.
- Work professionally in the language: this is the FSI target, and the big numbers apply.
The variables that move your timeline most
Two people can log identical hours and end up in different places. Talent is rarely the reason.
- Distance from languages you know. The biggest factor by far, and the entire basis of the FSI categories. Shared vocabulary buys you thousands of words for free.
- How the hours are distributed. An hour a day beats seven hours on Sunday. That’s the spacing effect again, operating at the scale of your calendar.
- How much of it is retrieval. Hours spent recalling and producing do more than hours spent passively reviewing.
- Whether you speak early. People who wait until they feel ready take longer to get comfortable. The waiting is the problem.
- Whether you continue at all. The dominant term, and it isn’t close. Most people who fail to learn a language didn’t learn slowly. They stopped.
A more honest way to estimate
If you want a number for planning, build it yourself:
- Find your language’s FSI category for a rough difficulty multiplier.
- Pick a concrete goal, whether a conversation, a novel or a job, not "fluency."
- For conversational comfort in a Category I language, budget somewhere in the low hundreds of hours; scale up sharply for Category III and IV.
- Divide by the hours you will genuinely do each week, not your optimistic figure.
- Add a margin, because you will miss weeks.
Then ignore the total and watch the checkpoints. Month one is dramatic and obvious. Month six is slow, invisible, and every bit as real. That’s exactly where people quit, because a countdown gives you nothing to see.
Our CEFR-aligned learning paths are structured around exactly those checkpoints, so you can see which level you are working through instead of estimating against a number that was never about you.
The honest answer: longer than the apps imply, shorter than 2,200 hours for most of what you want, and decided almost entirely by whether you’re still at it in month eight.
