How to remember what you read

You finish a book, and six months later you can name the title and almost nothing else. The fix is not reading more carefully, it is what you do in the five minutes after each chapter.

Most reading is not meant to be remembered, and that is fine. This is about the other kind: the book you will want to argue about in a year, the paper your work depends on, the chapter you will be examined on.

For that kind, the ordinary approach fails in a specific and well-understood way. You read, you highlight, you feel you have it, and three months later you retain the impression of having read it and almost none of the content. Highlighting produces familiarity, and familiarity is what convinces you that you know something you cannot actually produce.

The short version

1. Read normally. Do not take notes as you go.
2. At the end of a chapter, shut the book and write down what you can remember.
3. Turn three or four of those into questions with a definite answer.
4. Review those questions on a widening schedule.

That is the whole method. Everything below is why each step is shaped that way.

Why not take notes while reading

Because copying is not remembering. A note taken while your eyes are on the sentence is transcription, and it runs on the page rather than on your memory. Closing the book first forces retrieval, which is the part that actually strengthens anything, and it has the side effect of showing you honestly how much of the chapter survived. It is usually less than you expect, and knowing that is useful.

Turning it into questions worth keeping

A good question has one answer you could say out loud, and it makes sense on its own in six months when the context is gone.

Three or four per chapter is the right order of magnitude. This will feel too few. Keep in mind that each one is a small recurring commitment, and forty questions from one book is a chore you will abandon, which leaves you with nothing rather than with four.

Choosing what is worth keeping

Before writing a question, ask what you would lose by forgetting it. A number you could look up in ten seconds is usually not worth it. Something that changes how you think about a subject, or that you need to have in your head to follow an argument, usually is.

Anything you can look up quickly and rarely need is fine to look up. The point of keeping something in memory is not completeness: it is being able to think with it in the moment, when there is no time to search for it.

The review itself

Answer before you check. The half second of discomfort when you cannot quite reach it is the part doing the work, and skipping to the answer turns the whole exercise back into rereading.

Be honest when you did not know it. Marking something as recalled when you actually half-guessed it pushes it into a longer gap, and you will lose it. Nothing is being scored, and the only person a lenient mark harms is you.

A schedule that widens is the part people cannot do by hand. Three or four questions a chapter, across a few books, across a year, is hundreds of separate due dates. That is exactly the kind of bookkeeping to hand to software.

Also worth reading: what spaced repetition is and why it works and how this compares to Anki.