The hardest languages to learn for English speakers, ranked

The hardest languages for an English speaker are Arabic, Cantonese, Mandarin, Japanese and Korean. That is not an opinion: it is the top category of the best-known public ranking built on teaching adults to a fixed standard, the one kept by the US Foreign Service Institute. What that ranking cannot tell you is why those languages take so long, so this post adds a second measurement — how many words it takes to understand ordinary speech in each language — and puts the two side by side.

Updated 25 September 2026 · 6 min read

The ranking: four categories of time

The Foreign Service Institute (FSI) trains American diplomats, and it publishes how long its students usually take to reach an integrated score of 3 (Speaking + Listening) on the ILR scale. It sorts languages into four categories by that time:

So the ranking, as far as the FSI goes: Category IV languages take about three times as long as Category I, and nearly four times as long as its fastest languages. The FSI calls them "super-hard languages". Category III, which it calls "hard", is the long middle — the page notes that its list is not exhaustive — and it contains most of the world's major languages that are not close relatives of English.

Within a category the FSI does not rank, and neither should anyone quoting it: Japanese is not "harder than Korean" on this evidence, only in the same band. The weeks are also a measure of its students — adults selected partly for aptitude, studying full-time — not of you. What transfers is the ratio.

Why: the vocabulary does not arrive as fast

Part of the difficulty is obvious — a new writing system, tones in Mandarin and Cantonese, almost no vocabulary shared with English. Part of it can be measured. Take a large corpus of each language's film and television dialogue (OpenSubtitles 2018, close to how people actually talk), count how often every word form occurs, and ask how many of the most frequent forms it takes to cover a given share of what is said. Every corpus is first sampled to the same size, 403,882 words, so the comparison is fair.

Mandarin's 10,000 most common words cover 88.6% of its dialogue. Here is how many words each of the other languages needs to reach the same point:

Spanish gets there with 5,607 words — roughly half what Mandarin needs — and French with 3,609. The whole curve looks like this:

Mandarin's curve is the flattest of the seven at every point. Part of that is real — Chinese makes new words by compounding freely, and a great many of them are only moderately common — and part is measurement: written Chinese has no spaces, so where one word ends is decided by software, and a different segmenter would move the curve. Counting characters instead of words gives an almost vertical curve; both pictures are explained in how many words you need to speak Mandarin.

The seven languages we measured, ranked

Putting the two measures together for the languages we could measure gives an honest ranking, hardest first:

1. Mandarin — FSI Category IV, 88 weeks, and the flattest vocabulary curve: 12,141 words for 90% of speech. On top of that, a writing system that has to be learned separately.
2. German — Category II, 36 weeks. Its vocabulary curve is actually steep at first (77.6% with 1,000 words); the difficulty is in the grammar — cases, genders and word order — and in a long tail of compounds. See how many words you need to speak German.
3. Spanish and French — Category I at 30 weeks, the upper end of the easy band. French has the steepest curve of all (79.5% with 1,000 words) but a spelling that hides its pronunciation; Spanish is the reverse.
4. Italian and Portuguese — Category I at 24 weeks, the fastest the FSI lists.

Esperanto was never rated by the FSI, and Shanghainese is not on its list at all; the first is covered in the easiest languages for English speakers, the second has its own guide to how long it takes.

What "hardest" should mean to you

The FSI's categories measure the time to a high professional level. The vocabulary curves measure something that matters sooner: how quickly the words you know start covering what you hear. A Category I language rewards the first thousand words with a lot of understanding; a Category IV language rewards the same effort with noticeably less, and asks for a writing system on top.

Neither makes a language impossible, and "hardest" is a statement about time, not ability. If you are choosing, the steepest-frequency-curve comparison has every figure for every language, and each language's word list — Spanish, French, German, Mandarin — starts with the words that pay back fastest.

The Spanish, French and German lists can also be downloaded free: Spanish, French and German as printable PDFs, or Spanish, French and German as CSV files for Excel or Anki.

Sources

• Foreign Service Institute, National Foreign Affairs Training Center: language learning timelines, checked 2026-09-25.
• Lison, P. & Tiedemann, J. (2016). OpenSubtitles2016: Extracting Large Parallel Corpora from Movie and TV Subtitles. Proceedings of LREC 2016. aclanthology.org
• Dave, H. FrequencyWords: frequency lists from OpenSubtitles 2018, content licensed CC BY-SA 4.0. github.com. Every coverage figure is computed by scripts/frequency_curves.py and printed from its output rather than typed.