Write down the things you could not rebuild later, and nothing else. A name, a figure, a date and a proper noun are gone the moment they pass, because nothing else in the talk implies them. An argument is not: give yourself four nouns and the connections between them come back on their own.
That is the whole selection rule, and it exists because the writing is not free. Forming a phrase on paper uses the attention that was running ahead of the speaker, so every clause you write costs you the sentence after it. Notes are not a record taken alongside listening. They are taken out of it.
Eight rules for taking notes in a second language
- do not transcribe. Writing everything means understanding nothing, and the recording already exists.
- listen for the signposts: firstly, however, the key point is. These tell you what is coming and what matters.
- write nouns, not sentences. Content words carry the meaning and can be reconstructed later.
- abbreviate ruthlessly, dropping vowels and endings, since the notes only have to be readable by you within a day.
- use symbols for common words, an arrow for leads to, a plus for and, a question mark for a doubt.
- leave the left margin for structure, marking main points and sub points as you go rather than afterwards.
- one page per talk forces selection, which is the whole skill.
- review within a day, filling gaps while the memory can still supply them. Notes reviewed a week later are usually unusable.
Decide the shorthand before the talk, never during it
Inventing an abbreviation in the moment costs exactly what writing a sentence costs, and for the same reason. So the symbols have to be settled in advance and reused until they are automatic: an arrow for a consequence, a plus for an addition, a question mark for something to check, and three or four abbreviations for the words this particular subject will repeat forty times.
Fix them once and never revise them mid-session. A note-taker who is choosing between two abbreviations has stopped listening as completely as one writing a full clause, and the choosing feels like efficiency, which is what makes it hard to notice.
The two minutes after are part of the technique
The notes are deliberately incomplete, so they have a shelf life measured in minutes. Leave the room and the nouns are still there and the connections are not, and by the evening you have a list of words with nothing joining them.
So the reconstruction has to happen immediately: two minutes, on the spot, writing the connections between the nouns while the memory of them is still available. That is not tidying up. It is the step that turns the notes into a record, and it is the step people skip because the talk is over and leaving feels like the natural next move.
Where it stops: writing sentences
The stall is taking good notes. A learner writes clauses, because a clause is what a note looks like in their own language and because incomplete notes feel like poor ones. The pages that come out are legible, orderly and often quite full, and comprehension during the talk has collapsed to a fraction of what it was.
What hides it is the notes themselves. They read well afterwards, so the session is judged a success, and the twenty minutes of the talk that went past unheard while a sentence was being formed leave no trace at all. Nothing in the artefact records what it cost.
The test is uncomfortable and quick. Listen to ten minutes taking no notes whatsoever, then write what you remember. Most people find they have more than the notes would have given them, which settles the argument about what the writing was buying.
A layout that leaves room for the rebuilding
The Cornell layout is the practical answer and it is free, which is to say it is a line drawn down a page. A narrow column on one side, a wide one for the notes, a strip along the bottom. The nouns and numbers go in the wide column during the talk; the narrow column and the strip are filled in during the two minutes afterwards, with the connections and one sentence saying what the talk was for.
It replaced writing continuous prose down the page, which cannot be reviewed because there is nowhere to put the review. Walter Pauk set it out at Cornell in the nineteen fifties and it has not needed changing since, which is unusual and is a point in its favour. Its limitation is that it assumes you get the two minutes: in an exam listening section you do not, and there the strip stays empty and the notes have to survive on the nouns alone.
What it looks like when it goes wrong
The recognisable failure is a page of notes with a hole in the middle. You wrote a long point down, and the next two minutes of the talk are simply absent, and the notes give no sign of it because they continue smoothly on the other side.
The second failure is notes in your own language. Translating while listening is the most expensive thing you can do with the attention, more expensive than writing English clauses, and the resulting page is a record of your translation rather than of the talk.
The prediction that the writing is competing with is described in real speed, real accents, no subtitles, and the material where all of this is hardest, because nothing is repeated and there is no picture to fall back on, is listening with no pictures to lean on.
What this leaves open is what happens to the notes next. A page of nouns and arrows is an input rather than an output, and turning it into something a reader can follow is a different skill running under different rules: one idea, an opening line, and how the next one connects.