Listening is often the time sink — Latin shows why

Ask a learner what is hardest and they will usually say grammar or vocabulary. Ask someone at B1 why they cannot follow a film and the answer is different: everything arrives too fast and does not stop.

There is a clean way to see how much of a language is listening, and it is not a study. It is Latin.

The natural experiment

Latin is learned almost entirely for reading. The largest Latin exam in the world tests reading and written comprehension. Almost nobody's goal involves understanding spoken Latin at conversational speed.

Take listening and speaking out of a language and look at what happens to the structure of the problem:

Immersion disappears as a method. In our planning model, living in the language reaches an estimated 0.75× method factor at C1 — the only method that gets better at higher levels. C1 here is our planning scale, not NFATC's published endpoint. For Latin there is nowhere to go. The best method is unavailable, not slower.

Tutoring loses most of its advantage. A tutor's value is real-time correction of what you produce. If you are not producing spoken sentences, the largest part of that is gone. What remains — a commentary, a teacher explaining a construction — is closer to a good book.

Self-study stops having a ceiling. On our estimated CEFR planning scale, the model caps structured self-study at C1 and app-only at B1, because the levels above B1 are made of corrected novel output. Reading has no such requirement. People reach genuine expertise in reading Latin alone, which essentially nobody does in spoken Mandarin.

Three structural features of language learning, all of them consequences of listening and speaking, all of them absent at once.

What that implies about the hours

If removing listening removes the best method, most of the tutoring advantage, and the ceiling on self-study, then listening is not one skill among four. It is the constraint the others are organised around.

I want to be careful here, because this is inference rather than measurement. No published source apportions the current NFATC benchmark between Speaking and Listening, much less across all four broad skills, and any specific split — "40% listening" — would be invented. What the Latin case supports is the structural claim, not a percentage.

The structural claim is enough to act on.

What follows for a spoken language

Listening practice is not optional and it is not passive. The failure mode is comprehensible-input-shaped: hours of audio at a level far above yours, which is pleasant and teaches little. Material you can follow at 80% is where the work happens.

It explains the B1 wall better than grammar does. At B1 you can construct sentences and cannot follow a conversation between two native speakers. Those are different skills and the second one has had far less of your time.

It is why immersion improves with level while everything else degrades. Immersion is unlimited unpredictable input. That is worth little when you know two hundred words and enormous when you know five thousand.

And if your goal really is reading

Say so, and plan differently. A reading-only target in a modern language is a different project from the current NFATC benchmark, which targets integrated ILR 3 in Speaking and Listening.

Nothing on this site models that. We map NFATC's integrated oral benchmark onto a CEFR curve for planning, with C1 as an explicit site estimate rather than an official crosswalk. If you only want to read Russian novels, the figure on the Russian page is not a direct estimate of your project, and I cannot convert it without inventing a number.