Before deciding that something is at the right level, find out what is doing the carrying. A video you follow comfortably may be comprehensible because of the faces, the room and the situation rather than because of the words, and the same speaker on audio alone will tell you within a minute which it was.
That test matters because material only teaches at the edge of what you can do, and support quietly moves the edge without moving you. An hour that felt like the sweet spot and was actually being carried by the pictures has taught you less than the comprehension suggested.
Eight rules for choosing what to consume
- understood plus a little, which is the whole principle and everything below is a way of hitting it.
- one unknown word in fifty for fluency work, where you never stop and the goal is volume.
- one in twenty for study, where you do stop, look up and take notes. Confusing these two modes is the commonest mistake.
- narrow input beats varied input. One author, one podcast, one topic for a month, because the repeated vocabulary is what fixes it.
- repeat the same material two or three times rather than always moving on.
- subtitles in the target language only. Subtitles in your own language turn the exercise into reading your own language.
- quantity matters more than perfection. Understanding eighty per cent of a lot beats understanding all of very little.
- measure hours, not pages. Time is the honest unit and it is the one that predicts progress.
The idea, and the year it was named
Stephen Krashen set this out in 1982 as input just beyond the current level: material you can follow with the help of context, containing things you do not yet have. The part that gets quoted is the beyond, and the part that decides everything is the with the help of context, because context is exactly what varies between a video and a podcast of the same conversation.
The honest reading is that comprehensibility is a property of the pairing rather than of the material. The same recording is noise for one learner, the sweet spot for another and already easy for a third, and none of that is written on it.
How to tell, in five minutes
Take five minutes of something and play it once, with no stopping, no rewinding and nothing looked up. Then write down what happened, in whatever language you like.
If you can write what happened and there were words in there you did not have, that is the band you want. If you can write what happened and nothing in it was new, the material is behind you however much you enjoy it. If you cannot write what happened, it was noise, and no amount of persistence converts noise into input.
The measure is what happened rather than a percentage of words, because a percentage cannot be estimated honestly from inside and the summary can be checked.
Where it stops: mistaking support for comprehension
The stall is a learner who has calibrated on video. Following a series comfortably, they conclude they are at that level, and choose everything else at the same difficulty. Then a phone call, a podcast, or somebody talking in the next room turns out to be impossible, and the drop looks like a listening problem specific to those situations.
It is not specific to them. What has happened is that the comprehension was a joint product of the language and the pictures, and the pictures were doing more of it than anybody could see. The learner has been practising at a level below their own for months while believing the opposite.
Nothing corrects it, because the video keeps working. Success on the material you chose is the least informative signal available, and it is the only one most learners ever collect.
Checking what you got against what happened
Wikipedia's plot summaries are free, need no account, and exist for almost every film and for the episodes of most series. Watch, write down what you think happened, then read the summary and compare. It is the cheapest external check available and it costs two minutes.
It replaced asking somebody who has seen it whether you understood it correctly, which requires that person to exist and to be patient.
Its limitation is granularity. A summary confirms the shape of the plot and says nothing about the sentences you missed, so it catches the case where you followed nothing and thought you had, and not the case where you missed a joke, a motive or a qualification. For those you need the transcript rather than the summary.
What it looks like when it goes wrong
The recognisable failure is a learner who follows films and cannot follow a phone call. Almost every hour they have logged came with a face attached, and a voice with no face is a different task they have not practised.
The second is material chosen for comfort. Comprehension at ninety percent feels like the right level and is usually a signal to move, because the ten percent left is not enough to be worth an hour.
The second condition, which is whether you were attending at all, and the reason most logged hours fail it rather than the difficulty test, is filling your day with the language without moving country, and the routes that a word needs before it arrives on its own are described in the curve everyone is fighting, and what bends it.
What none of this touches is production. Every hour discussed here builds comprehension, and a year of well-chosen input produces somebody who follows a film and freezes in a conversation, which is a different problem with a different fix: finding someone to speak to, and what to do in the hour.